AI Search Statistics & Trends 2026

AI Search went mass-market in 2026 while the click kept shrinking. ChatGPT reached 1.2 billion weekly users and Google's AI Mode passed 1 billion monthly users, yet 68% of US Google searches now end without a click.

This section collects the most useful AI search statistics published between January 1 and September 30, 2026. Each entry gives you the stat in one line, a link to the primary source, and a short read on what it means for your strategy. If a 2026 roundup re-quoted a 2025 study, we left it out.

One warning before you dig in: panels, rank trackers, first-party analytics, and surveys all measure different things. When two numbers disagree, the context tells you why. Use the direction of the data with confidence and the exact figures with care.

AI Search Adoption and Usage

AI search is now a mass-market behavior, and it's stacking on top of Google rather than replacing it.

Gemini's monthly web visits grew 548% in 2025, compared with 43% for ChatGPT. (Similarweb, January 2026)

gemini.google.com went from 267.7 million visits in January 2025 to 1.74 billion in December, while chatgpt.com went from 3.85 billion to 5.52 billion after peaking at 6.17 billion in October. Claude grew 125% and Perplexity 80% over the same year. The challengers are where the growth is now.

ChatGPT's search usage nearly tripled in a year. (OpenAI, March 2026)

OpenAI shared this alongside its $122 billion raise, noting ChatGPT draws six times the monthly web visits and mobile sessions of the next largest AI app. There's no baseline, so treat it as direction rather than a hard number. The direction is the point: the retrieval layer inside ChatGPT is where your content gets found or skipped.

Google accounted for 73.7% of US desktop search activity across 41 major search-capable sites in Q4 2025, and AI tools accounted for 3.2%. (SparkToro, March 2026)

Amazon, Bing, and YouTube each still see more desktop search activity than ChatGPT. Desktop-only clickstream likely undercounts mobile-first AI apps, but the takeaway holds: search happens everywhere, and AI is one slice of it. Don't walk away from the channels that still carry most of the demand.

ChatGPT ran a live web search on 34.5% of US prompts in February 2026, down from 46% in late 2024. (Semrush, April 2026)

A smaller share of prompts triggers search, even as OpenAI says total search usage is climbing. Prompts that do trigger search nearly doubled in length, from 4.7 to 8.7 words, and Google now receives more than 21% of ChatGPT's outbound referral clicks. Your ChatGPT visibility and your Google visibility are more connected than the platform wars suggest.

Google says Search queries are at an all-time high, and that AI Mode is driving an incremental increase in queries overall. (Alphabet Q1 2026, April 2026; Alphabet Q2 2026, July 2026)

Search usage also set a record during the 2026 World Cup. Google doesn't publish absolute query counts, so you're taking its word on the trend. Still, anyone calling search dead has to explain record usage. Search isn't dying. It's changing shape.

Bing reached 1 billion monthly active users for the first time. (Microsoft FY26 Q3 earnings, April 2026)

Bing punches above its market share because its index powers Copilot. If Bing Webmaster Tools has been an afterthought, fix that. A second index is a second shot at being retrieved.

31.3% of the US population will use generative AI search in 2026. (EMARKETER, April 2026)

This is a forecast, not a measurement. EMARKETER also finds people under 44 average five search platforms. AI search is one more place your audience looks, not the only one.

49% of US adults now use AI chatbots, up from 33% in 2024, and about a quarter use them daily. (Pew Research Center, June 2026)

ChatGPT leads in every age group, with Gemini a distant second. Surveys that ask differently run higher: Elon University found 65% of US adults used an LLM in the past week. However you slice it, half the country is already here.

42% of US adults use chatbots to search for information, the top chatbot use case. (Pew Research Center, June 2026)

That climbs to 54% for adults 18 to 29 and drops to 20% for those 65 and older. Information search beats work tasks, entertainment, and news. When people open a chatbot they're usually looking something up, which makes it a search engine whether it calls itself one or not.

The Gemini app has 950 million monthly active users, and its daily users tripled in a year. (Alphabet Q2 2026 earnings, July 2026)

Gemini sat at 750 million monthly users in February 2026 and 900 million-plus at I/O in May. Google is converting Android, Chrome, and Workspace distribution into AI usage faster than anyone can buy it. That's the Google advantage playing out in public.

ChatGPT's share of worldwide generative AI web traffic fell from about 76% to about 53% in a year. (Similarweb, July 2026)

Between June 2025 and May 2026, Gemini climbed from under 9% to roughly 27%, and Claude from about 2% to nearly 9%. ChatGPT is still the biggest room, but it's no longer the only one that matters. Track your visibility across engines, because the audience is spreading out.

71% of Claude conversations are work-related, compared with about 40% on ChatGPT and 43% on Gemini. (Profound, July 2026)

The panel covers 100,000 conversations per platform from July 2025 to June 2026, labeled with OpenAI's own usage taxonomy, though Claude coverage spans only nine countries. On weekdays, programming leads on Claude (about 35%), writing on ChatGPT (about 28%), and multimedia generation on Gemini (about 29%). Each assistant does a different job, so prioritize engines by what your buyers use them for.

Visits to generative AI platforms hit 9.5 billion a month in May 2026, up 70% year over year, yet 95% of ChatGPT users still use Google. (Similarweb, August 2026)

That 95% overlap hasn't moved since September 2025. People aren't leaving Google for ChatGPT. They're using both, for different jobs, and your strategy has to cover the whole journey.

ChatGPT now reaches 1.2 billion weekly users, according to OpenAI. (OpenAI, September 2026)

That's up from 800 million in October 2025 and 900 million-plus in March 2026, roughly a third more users in six months. At that scale ChatGPT isn't an emerging channel. It's a search surface with more weekly reach than almost anything on the web, so plan for it the way you plan for Google.

25% of Americans use AI every day, up from 19% a year earlier. (Menlo Ventures, September 2026)

Overall adoption only moved from 61% to 64%, so the growth is existing users going deeper. Gemini's reach among AI users jumped from 36% to 58%, nearly matching ChatGPT's 60%, and Claude's nearly tripled to 20%. The land grab for new users is slowing. The fight for daily habits is on.

ChatGPT's US app stickiness averages 40%, roughly double any other AI app. (Similarweb, September 2026)

Stickiness is the share of monthly users who show up on a given day, and Claude averaged 21% and Gemini 20% on iOS and Android over the year. ChatGPT doesn't just have more users. Its users come back more often.

Claude's six-month user retention rose from 25.8% for its January 2025 cohort to 44.8% for its February 2026 cohort. (Similarweb, September 2026)

ChatGPT's comparable figure is 59%. Over the 12 months through August 2026, Claude's share of visits across six major chatbot sites rose from 2% to 9.6%, and the share of ChatGPT users who also used Claude went from 2% to 14%. Your audience is going multi-assistant, so visibility on one engine isn't coverage.

The median prompt is 17 words on Claude, 13 on ChatGPT, 12 on Gemini, and 9 on Google AI Mode, versus 3 words for a traditional Google search. (Similarweb, September 2026)

That's US desktop data from June 2025 to May 2026. Claude users also send the most prompts per session (5.6), followed by ChatGPT (5.0). Keyword research built on three-word queries can't describe what people are actually asking these systems.

↑ Jump to Top

Google AI Overviews and AI Mode

Google's AI answers now reach billions of people and sit on somewhere between a third and half of all searches, and they're moving into commercial queries.

AI Mode queries are three times longer than traditional searches, and nearly 1 in 6 is voice or image. (Alphabet Q4 2025 earnings, February 2026)

In the US, daily AI Mode queries per user doubled since launch, and a significant share of queries lead to a follow-up question. Long, multimodal, multi-turn queries break keyword-level tracking. This is the move from keywords to conversations, and it's already here.

The average AI Overview is now more than 1,200 pixels tall, taller than a standard desktop screen. (BrightEdge, February 2026)

Height grew about 15% in a year and peaked near 1,340 pixels in December 2025. When the answer fills the viewport, position one lives below the fold. Being inside the answer is the new above the fold.

After Gemini 3 came to AI Overviews, about 42.4% of previously cited domains no longer appeared, and sources per AI Overview rose from 11.55 to 15.22. (SE Ranking, February 2026)

The study covers 100,000 US keywords. A model upgrade reshuffled nearly half the cited domains, almost all of them long-tail sites, while only one of the 500 most-cited domains dropped out. Model launches are the new core updates, so watch release notes the way you watch Search Central.

AI Overview coverage nearly doubled from 23% to 47% of US searches between September 2025 and January 2026, then fell back to 34% in February. (Conductor, April 2026)

The study covers 274.5 million searches across 11 industries. Six-month averages run from 46.4% in Communication Services and 42.8% in Health Care down to 20% in Real Estate. Treat any single month's prevalence number as a snapshot, because Google turns this dial constantly.

AI Overviews appear on 36% of informational queries but only 8% of commercial and 5% of transactional queries. (Seer Interactive, April 2026)

Comparison queries trigger AI Overviews 95.4% of the time and question-style queries 85.9%, based on 5.47 million queries across 53 brands. If your content answers questions or compares options, assume Google is summarizing it before anyone clicks.

AI Overviews has more than 2.5 billion monthly active users. (Google I/O 2026 keynote, May 2026)

Google doesn't say how it counts a monthly user, and AI Overviews appear without anyone opting in, so read this as exposure, not engagement. It's still the largest AI answer surface on the planet by a wide margin. If you only track ChatGPT, you're missing the AI answer your customers see most.

AI Mode passed 1 billion monthly users within a year of launch, and AI Mode queries have more than doubled every quarter. (Google, May 2026)

At I/O, Google also merged AI Overviews and AI Mode into one experience and made Gemini 3.5 Flash the default model in AI Mode. The merge makes the AI Mode user count harder to interpret. It also tells you the line between search and AI search at Google is gone.

AI Overviews appeared on 86.7% of Peec AI's sampled business-intent searches in April 2026, up from 56.9% a year earlier. (Peec AI, May 2026)

The sample is 500,000 prompts that lean toward buying intent and exclude navigational queries, which helps explain why it runs far above the panel numbers earlier in this section. Coverage was 91.4% in the US and 91.7% in the UK, 76% across the EU, and zero in France, which didn't get AI Overviews until July. For the commercial questions your buyers type, an AI Overview is close to guaranteed.

Two-word queries triggered an AI Overview 64.6% of the time, compared with 89.1% for queries of 11 to 15 words. (Peec AI, May 2026)

Five-word queries hit 78.5% and eight-to-ten-word queries 87.7% before the rate plateaus. Queries with a clear intent shape triggered AI Overviews 93% to 97% of the time. Past the shortest keyword queries, the longer and more specific the question, the more likely Google answers it itself.

Only 0.34% of US Google searches went to AI Mode from January through April 2026. (SparkToro, June 2026)

That sits awkwardly next to 1 billion monthly users, and the gap is likely definitional. Google reports monthly users without saying how it counts them, while SparkToro's Similarweb clickstream counts searches that land in AI Mode and excludes the Google app. AI Mode has reach, but classic results topped by AI Overviews still carry most of the volume.

AI Overviews on commercial-intent SERPs grew 71% in six months. (Semrush, July 2026)

Across 600,000-plus keywords, finance commercial queries with AI Overviews rose 231%, and ads and AI Overviews now share a SERP about twice as often as a year earlier. The informational safe zone is over. AI Overviews are moving down the funnel toward the money.

Text ads appear on 29.45% of commercial AI Mode queries. (SE Ranking, July 2026)

Ad presence climbs with CPC, from about 24% on keywords under $2 to about 54% at $10 and up. Only 11.53% of advertiser domains were also cited as sources in those answers. Paid and earned visibility in AI Mode are separate games, and you need a plan for both.

Keywords with AI Overviews carry much higher CPCs in expensive categories: $5.02 versus $1.51 in Jobs and Education, and $4.84 versus $2.14 in Finance. (Semrush, July 2026)

The pattern reverses in lower-cost industries like internet and telecom, games, and food and drink. More than 60% of the most expensive Jobs and Education keywords showed an AI Overview. Google is putting AI answers where clicks are worth the most, which is exactly where you can least afford to be missing.

google.com's citation share in AI Mode rose 8.4x between April 15 and June 30, 2026, making it AI Mode's #2 cited domain. (Profound, July 2026)

The growth came almost entirely from Business Profile and Product Knowledge Panel pages, and it hit hospitality, travel, home services, restaurants, real estate, and healthcare hardest. SE Ranking measured google.com at 17.42% of AI Mode citations in February, up from 5.7% in June 2025. For local and product queries, your Business Profile and product data are citation sources now.

39.4% of US Google desktop searches included an AI Overview in June 2026, up from 25.8% in July 2025. (Comscore, September 2026)

This is an independent panel measure, not a rank tracker's keyword set. Rank trackers run higher: BrightEdge saw AI Overviews on 48% of its tracked queries in February 2026. Either way, you should assume an AI answer sits on top of a big chunk of your keyword set.

AI Overviews grew from 38% to 57% of tracked manufacturing search volume between January and July 2026. (Semrush, September 2026)

The AI Overview figures come from 458 manufacturing and industrial keywords. Separately, Semrush clickstream data showed AI tools sent just 0.48% of US sessions to manufacturing sites, and only two domains (Vevor and Grainger) made both the top 15 most-mentioned and most-cited lists. Even the least AI-native B2B categories are getting summarized now.

Google AI Mode's citations fell 82% in three weeks, and by early September 3.5 of the 3.8 remaining sources per answer were google.com/goto redirect links. (Evertune, September 2026)

In mid-August a typical AI Mode response carried 19.7 citation URLs and 5 plain links, based on about 180,000 responses. Gemini's sources fell about 30% and AI Overviews' about 15% over the same stretch. Google blamed a model bug, but Evertune says the decline started earlier and tracks the redirect rollout. If your AI Mode referral data shifted in September, start here.

↑ Jump to Top

Clicks, CTR, and zero-click

Every independent method points the same way: AI answers cut organic clicks roughly in half at the top of the page, and being cited is the only way to claw some back.

A top-ranking page's click-through rate is 58% lower when an AI Overview is present. (Ahrefs, February 2026)

The 2025 edition of the same study found 34.5%, so the gap has widened. The study compares Search Console data for 300,000 keywords, December 2023 against December 2025, and the hit shrinks further down the page, from about 51% at position two to 19% at position ten. Ranking first still matters. It's just worth less than half as much when an AI Overview shows up.

In Germany, clicks on the #1 organic result fall from 27% to 11% when an AI Overview appears. (SISTRIX, February 2026)

Across 100 million-plus keywords, SISTRIX estimates AI Overviews cost 265 million organic clicks a month across all positions, or 6.6% of all organic clicks. It's the same pattern Ahrefs and AWR found in the US, measured in a market with different rules. The CTR hit isn't a US quirk.

On informational queries, being cited in an AI Overview earns 120% more organic clicks than not being cited, but still 38% fewer than when no AI Overview appears. (Seer Interactive, April 2026)

Per million impressions on informational queries, Seer measured about 33,500 clicks with no AI Overview, about 20,700 when the brand is cited, and about 9,400 when it isn't. That's the business case for GEO in three numbers. You can't opt out of the AI Overview, but you can decide which side of that gap you land on.

Organic CTR on queries with AI Overviews rebounded 85%, from 1.3% in December 2025 to 2.4% in February 2026. (Seer Interactive, April 2026)

It's the first reversal after about 18 months of decline, and queries without AI Overviews improved too, from 3.2% to 3.8% over the same two months. Two months doesn't make a trend. Watch it, but don't bank on it.

Paid CTR on queries without an AI Overview fell from 26% to 21.8%, while paid CTR on queries with one rose from 14.6% to 16.2%. (Seer Interactive, April 2026)

That's January 2025 through February 2026 across 296.9 million paid impressions. Paid CTR is still much lower when an AI Overview shows, but the gap is narrowing. If you run paid search, your CTR benchmarks are moving too.

68% of US Google searches ended without a click from January through April 2026, up from 60.45% in 2024. (SparkToro, June 2026)

Fewer than one in three searches now sends a click anywhere. The 2024 baseline came from a different clickstream panel, so read the gap as direction, and the study excludes the Google app, so the real rate is likely higher. The click is no longer the default outcome of a search. Set your KPIs accordingly.

US Google searches send just 231 clicks to the open web per 1,000 searches, and no country studied reaches 300. (SparkToro, June 2026)

Zero-click rates across the six markets run from 62.1% in Germany to 69.5% in the UK. The rest of those searches end on Google's own properties or with no click at all, which is why owning your brand's presence on Google surfaces matters as much as ranking.

Google says it sends "billions of clicks to websites every week through AI features in Search." (Alphabet Q2 2026 earnings, July 2026)

That's Google's main answer to the independent data, and it comes with no denominator or trend. The new Search Console generative AI performance report shows impressions from AI Overviews and AI Mode, not clicks. Until Google shows the click data, measure it yourself.

92.8% of ChatGPT product-comparison tasks in a think-aloud user study ended with no meaningful click to the open web. (Profound, July 2026)

Across 221 tasks recorded with 56 participants, the brands people chose appeared about twice as often in ChatGPT's answers as the ones they passed over (24% share of voice versus 11%). Only 3.2% of tasks involved checking a claim on an outside site. The sample skews toward experienced users, but the lesson travels: in AI shopping, the shortlist is the shelf.

In a randomized field experiment, an AI Mode-only Google cut click-through to external sites by 18.8 percentage points. (Wang et al., arXiv, August 2026)

Removing AI features entirely raised click-through by 8.8 points. The preregistered study of 1,100 US users also found AI Mode cut the share of users clicking through to Reddit (down 21.2 points), news sites (down 12.5), and Wikipedia (down 9.9). It's a preprint, not yet peer reviewed, but it's the strongest causal evidence so far: AI answers don't just coincide with fewer clicks, they cause them.

Position-one organic CTR on US desktop falls from 20.02% to 9.69% when an AI Overview is present, and 74.3% of AI Overview SERPs end without an organic click. (Advanced Web Ranking, September 2026)

Without an AI Overview, 53.6% of SERPs end with no organic click. Stack an AI Overview with a local pack, product grid, People Also Ask, and video, and AWR estimates the top result loses up to 99% of its clicks. These are modeled estimates from 1.75 million-plus US SERPs a month, not Search Console data, but they line up with everything else here.

The average #1 organic result lost about a quarter of its clicks in a year, falling from 28.0% to 20.9% average CTR. (Advanced Web Ranking, September 2026)

Desktop took the hit while mobile held up: AWR shows desktop #1 CTR dropping from 45.6% in February 2026 to 14.6% by May, while mobile rose from about 30% in December 2025 to 39% by June 2026. Segment your CTR reporting by device, or the blended number will mislead you.

CTR for the most exposed French sites fell 23.1% within nine days of AI Overviews launching in France. (Ahrefs, September 2026)

Across 963 French domains, median CTR fell 5.7% while impressions held steady, and sites lost about one point of CTR for every point of AI Overview exposure. France gave the industry a clean natural experiment. It confirms what the US data already told you.

↑ Jump to Top

AI referral traffic and conversion

AI referrals are still a small share of traffic, but they doubled in a year and now convert better than almost any other channel.

Google sends 190 times more traffic to websites than ChatGPT: nearly 40% of site traffic versus 0.21%. (Ahrefs, February 2026)

That's across 76,000 websites in Ahrefs Web Analytics. Ahrefs separately found all AI chatbots combined sent 0.28% of total web traffic in March 2026. AI is where the conversation is growing. Google is still where the visits come from.

AI referral traffic grew 66% in 2025 but still made up only 0.14% of all web visits. (Semrush, April 2026)

The dataset covers billions of visits to 50,000-plus sites across 17 industries. Google AI Mode traffic grew from 1,600 visits to 38.2 million by December 2025, still just 0.01% of the total, while organic search held 16.04% of traffic and declined in 13 of 17 industries. The fastest-growing channels are still among the smallest.

Referral traffic from ChatGPT to other websites grew 206% year over year, even as ChatGPT's own traffic leveled off. (Semrush, April 2026)

Comparing January 2025 with January 2026 in US clickstream data, outbound clicks roughly tripled while chatgpt.com plateaued near 1 billion monthly visits from November 2025 on. For most of the period, 65% to 85% of prompts couldn't be matched to any traditional search keyword. ChatGPT is sending more people outward, even if you can't map the prompts that got them there.

On May 7, 2026, the share of ChatGPT responses containing a link jumped from about 4.5% to 20% to 24%, and OpenAI referrals to monitored brand sites rose 60% to 65%. (Profound, May 2026)

It's the same homepage-link change Similarweb tracked above. B2B SaaS referrals rose more than 200% and fintech 60%, while ecommerce was essentially flat. Profound called it the largest single-day change in AI brand traffic it had measured all year, and it came from one product decision at OpenAI. Plan for platform changes you don't control.

AI-referred visitors to US retail sites converted 54% better than non-AI traffic in May 2026 and generated 53% more revenue per visit. (Adobe, June 2026)

Twelve months earlier, non-AI visits were worth 128% more. AI visitors were also 36% less likely to bounce, spent 53% more time on site, and viewed 23% more pages. The quality debate is over: AI traffic shows up having already done its research.

AI-referred traffic grew 194% year over year to travel sites, 138% to retail, and 105% to financial services in May 2026. (Adobe, June 2026)

Growth is slowing as the base gets bigger. In Adobe's Q1 2026 read, retail AI traffic was up 393% year over year. AI-referred travel visitors spent 70% more time on site and were 41% less likely to bounce. Decelerating triple-digit growth is still triple-digit growth.

AI platforms accounted for 0.32% of all website traffic from January to April 2026, up from 0.02% in 2024. (SE Ranking, June 2026)

Across 101,574 sites, ChatGPT sent 74.78% of AI referrals, followed by Gemini (11.56%), Perplexity (7.23%), Copilot (3.51%), and Claude (2.62%), with Claude up 320% year over year. In SE Ranking's January to April 2025 engagement data, AI visitors spent 68% more time on site than organic search visitors. The measured slice is small, but it's worth courting.

At n8n, GA4 last-touch credited about 1% of conversions to AI answers, while a post-conversion survey credited about 9%. (Graphite, June 2026)

Graphite found 90% of those AI-driven conversions never clicked a citation link, so they landed in GA4 as direct traffic, and they converted 33% better than organic search. It's one company, but it's the cleanest look yet at how badly last-touch undercounts AI. Ask your customers where they heard about you.

LLM referral sessions grew 9.9x in 19 months, yet they still make up only 0.02% to 1.71% of total sessions depending on industry. (Previsible, July 2026)

Across 166 GA4 properties, ChatGPT accounted for 92.4% of trackable LLM referrals from November 2024 through May 2026. Claude grew 64x and passed Perplexity in March 2026, while Perplexity fell 61% from its peak. Previsible also notes that AI discovery inside Google's own results almost certainly drives more AI traffic than every standalone assistant combined, and because AI Overviews don't produce trackable referral sessions, that traffic hides inside your Google organic numbers.

After ChatGPT's May 2026 homepage-link update, the share of its referrals landing on homepages jumped from under 30% to more than 60%. (Similarweb, July 2026)

ChatGPT switched to clickable brand names in answers instead of footnote citations, and the shift stuck. Similarweb says this traffic now behaves "less like a search click and more like the tail end of an ad impression." Page-level attribution from ChatGPT just got worse, so attribute at the query and entity level instead.

AI platforms sent an average of 770.7 million referral visits a month to websites worldwide between June 2025 and May 2026, up 117.4% year over year. (Similarweb, September 2026)

Marketplaces led at 46.8 million visits a month, with News and Travel tied at 44.5 million, while Beauty (up 312.5%) and Fashion (up 278.2%) grew fastest. Similarweb says AI referrals remain a low single-digit share of most sites' traffic but are compounding faster than any other channel it measures. Small today doesn't mean small next year.

ChatGPT drove a record 95.1% of AI referral traffic in August 2026, and its referrals grew 101% from January to August. (BrightEdge, September 2026)

Gemini had 2.4%, Perplexity 1.1%, and Claude 0.8%. That's a sharp swing from BrightEdge's own May report, where Gemini took 11.6% of AI referrals in Q1 2026. Referral shares move month to month, so don't build a plan on one platform's number. Gemini is winning users, but ChatGPT still sends the clicks.

People who received an AI recommendation for a brand were 2.5 times more likely to visit that brand's site within seven days, but only 8.8% of those visits came through an AI referral. (Similarweb, September 2026)

Search carried 55.9% of those AI-influenced visits, direct 19.9%, and other referrals 13.5%, based on US desktop journeys in Finance, Travel, and Beauty from July to December 2025. Profound found a similar lift (1.5x to 2.5x baseline visits), with more than 97% of those visits carrying no AI referral tag. Your AI influence is showing up in your search and direct numbers.

↑ Jump to Top

Citations and sources

Ranking on page one no longer guarantees a citation. Query fan-out, earned media, and freshness now decide which pages AI systems actually use.

When ChatGPT answered non-English prompts from matching countries, 43% of its fan-out searches ran in English. (Peec AI, February 2026)

The analysis covers 10 million-plus prompts and 20 million-plus fan-outs, and ChatGPT added English-language research in nearly 78% of cases. Turkish prompts triggered English research most (94%) and Spanish least (66%). Your English content is competing in markets you never targeted, and your local-language content is competing with it.

The average ChatGPT fan-out query roughly doubled in length, from about 6 words in October 2025 to about 12 by January 2026. (Peec AI, February 2026)

Fan-outs per prompt stayed flat at roughly 2.3 to 2.8, and the trend held across countries and languages. Engines also differ in how hard they search: Peec's April data had Grok at 6.8 fan-outs per query, versus 2.1 for ChatGPT and 1.4 for Perplexity. Sub-queries are getting more specific, which rewards pages that answer specific questions.

Only 17% to 38% of AI Overview citations come from pages ranking in Google's top 10. (Ahrefs, March 2026; BrightEdge, February 2026)

Ahrefs found 37.9% across 863,000 SERPs in its March 2026 update, down from about 76% in July 2025, with 31% of cited URLs ranking beyond position 100. BrightEdge's tracking from February to July 2025 puts top-10 overlap closer to 17%, and the spread comes down to different keyword sets and matching methods. Ahrefs credits query fan-out: the AI is ranking pages for sub-queries you never see. YouTube alone accounts for 5.6% of cited URLs, which is why video belongs in your plan.

ChatGPT left 85% of the pages it retrieved uncited, and 32.9% of the pages it did cite appeared only in results for a fan-out sub-query. (AirOps, March 2026)

Across 15,000 queries, pages ranking #1 on Google were cited 43.2% of the time versus 12.3% for pages beyond the top 20, and 95% of ChatGPT's fan-out queries had zero monthly search volume. Retrieval is the tryout. The citation is the roster spot, and your keyword tool can't see most of the plays that decide it.

LinkedIn jumped from about the 11th to the 5th most-cited domain on ChatGPT between November 2025 and February 2026. (Profound, March 2026)

It's now the top cited domain for professional queries across six AI platforms. Posts and long-form articles make up about 35% of LinkedIn citations, up from 27%, while profile citations fell from 33.9% to 14.5%. In B2B, what your people publish on LinkedIn is part of your AI search footprint.

Listicles account for 21.9% of all AI citations, more than any other content format. (Wix Studio AI Search Lab, March 2026)

Articles (16.7%) and product pages (13.7%) follow, across 1.06 million citations from ChatGPT, AI Mode, and Perplexity, and listicles make up about 40% of citations on commercial queries. Perplexity leans on discussion threads at more than double the rate of other engines. Structured, comparative content gets pulled into answers because it's easy to extract.

Distributing brand stories through third-party publishers produced a 239% median lift in AI citations. (Stacker with Scrunch, March 2026)

64% of the resulting citations came from third-party publishers rather than the brand's own site, across 87 stories, 30 brands, and eight AI platforms. Stacker sells distribution, so weigh this as vendor research. It still lines up with Muck Rack: AI engines weight what the web says about you.

About 11% of the listicle citations AI engines used for non-branded software-review prompts came from self-promotional listicles, where a brand ranks its own product. (Peec AI, March 2026)

ChatGPT's rate was 3.6%, versus 10.3% for Google AI Mode and 10.4% for Perplexity, across 232,000 citations over 12 weeks, and Peec found no evidence of algorithmic correction in that window. Self-promotional listicles still work on some engines, but ChatGPT cites them about a third as often. Don't build a strategy on a loophole.

Across 1.4 million prompts, cited pages had titles that matched ChatGPT's internal sub-queries more closely than the user's own prompt. (Ahrefs, April 2026)

Reddit URLs were retrieved in huge volume but cited just 1.93% of the time, making up 67.8% of all retrieved-but-uncited URLs. Ahrefs measured a higher overall cite rate than AirOps (about half of retrieved URLs), a reminder that method shapes the number. The lesson holds: write titles and headings that answer the sub-question directly. That's relevance engineering at the passage level.

91% of the search queries ChatGPT fires for a prompt are unique across runs, compared with 14% for Perplexity and 47% for Copilot. (Profound, April 2026)

Across 10,000 prompts over 14 days, all three engines fired about 1.4 to 2 searches per run, but ChatGPT almost never repeated a search string. On each prompt's most divergent query, word overlap with the original prompt was 88% for Perplexity and 13% for ChatGPT. You can't reverse-engineer ChatGPT's retrieval from one run, so simulate it at scale.

In ChatGPT, pages that were always cited had lower average domain authority (53 versus 56) and about a third of the backlinks (1.1 million versus 3.2 million) of pages that were never cited. (AirOps, April 2026)

In AirOps' March study, sites with DA 80 to 100 were cited for just 15.0% of their retrieved pages, versus 21.5% to 23.6% for every lower tier. Authority gets you retrieved. It doesn't get you cited. Relevance does.

ChatGPT cited pages whose best heading closely matched the query 41.0% of the time, versus roughly 29% to 31% for weak matches. (AirOps, April 2026)

Among strongly matched pages, those covering 26% to 50% of ChatGPT's fan-out subtopics were cited 38.2% of the time, beating pages that covered all of them (34.0%). Retrieval rank still dominated: a well-matched page ranked 11th or lower was cited only 21.5% of the time. Write focused sections that answer the sub-question, not exhaustive pages that try to answer everything.

ChatGPT cited pages of 500 to 999 words most often (34.3%), and pages over 5,000 words less often (28.6%) than pages under 500 words (30.5%). (AirOps, April 2026)

For articles, four to ten subheadings performed best (33.2%), while product pages did best with none (43.2%). Length isn't the lever in ChatGPT. Retrieval rank and query match are.

ChatGPT cited pages with JSON-LD schema 38.5% of the time, versus 32.0% without. (AirOps, April 2026)

Pages with FAQPage schema were cited 45.6% of the time, and AirOps says pages with and without schema were similar on length, headings, authority, and query match. It's still observational, though. Ahrefs' controlled test earlier in this section found no lift from adding schema, and Seer found the sites leaning hardest on FAQ and HowTo schema underperformed in AI Overviews. The schema debate isn't settled, so keep your markup clean and don't treat it as a silver bullet.

On awareness-stage commercial prompts, content with five to seven supporting statistics had a 20.3% higher likelihood of being cited by ChatGPT. (AirOps, April 2026)

On comparison prompts, pages with three tables earned 25.7% more citations and pages with seven distinct price points 15.7% more. These are model-based relative likelihoods across 217,508 retrieved pages, not percentage points. Specifics win: numbers, tables, and prices give the model something to extract.

ChatGPT cited pages 30 to 89 days old at the highest rate (32.8%), while pages under 30 days old had the lowest citation rate (25.3%). (AirOps, April 2026)

Only 41% of pages had a detectable publish date. Finance showed the strongest freshness effect (50.2% at 30 to 89 days versus 35.1% at five years or older), while ecommerce pages five years or older (36.7%) did about as well as those 30 to 89 days old (35.9%), and AirOps suggests brand-new pages may not be fully indexed yet. Fresh matters, but maintained beats brand new.

About 84% of AI citations go to earned media, and 27% go to journalism. (Muck Rack, May 2026)

Press releases got 1.1% and paid or advertorial content 0.3%, across 25 million-plus links from ChatGPT, Claude, and Gemini on realistic consumer prompts, some naming companies and some not. 57% of dated journalism citations were published in the past 12 months. AI leans on what others say about you far more than on what you say about yourself.

ChatGPT cites sources in 96% of its responses, Gemini in 82%, and Claude in 55%. (Muck Rack, May 2026)

ChatGPT averages about 5 citations per response and Gemini about 8, while Claude cites less often but averages 13 when it does. The top domains differ too: Wikipedia and Axios lead ChatGPT, PubMed Central leads Claude, and Reddit and YouTube lead Gemini. One content plan won't win every engine.

Adding schema markup to 1,885 pages produced no meaningful change in AI citations compared with 4,000 matched control pages. (Ahrefs, May 2026)

AI Overview citations dipped 4.6%, while AI Mode (up 2.4%) and ChatGPT (up 2.2%) moved within the noise. 53% of AI-cited pages use schema, but that correlation likely reflects overall technical quality rather than a direct effect. Keep your structured data clean for search features. Just don't expect it to move AI citations by itself.

Reddit captured 20.4% of all first-citation slots in Google AI Overviews, versus 1.94% combined for 14 major publishers. (Seer Interactive, May 2026)

The sample is 8,500 keywords from one week in May 2026. Seer calls the first citation the new position zero, and Reddit is winning it.

42.7% of the references ChatGPT cited in June 2026 were classified as AI-generated, up from 38.9% in January. (Graphite, June 2026)

Graphite ran a fixed set of information-seeking questions through logged-in ChatGPT and classified the cited pages with the GPTZero detector. The share was higher for entity prompts (45.8% in June) than editorial ones (27.5%), and both rose over the period. AI answers are increasingly built from AI-written pages. That's the content collapse happening in real time.

ChatGPT's minimal-reasoning and high-reasoning modes shared only 25.6% of cited domains for the same prompts. (Semrush, June 2026)

With high reasoning, sources per response rose from 2.6 to 4.5, Reddit fell from 15% to 7% of citations, and government and academic sources rose from 1.9% to 8.8%. The test covers only 100 prompts on GPT-5.2, so treat it as directional. Instant and Thinking modes behave like different engines, and your tracking should know which one it's sampling.

ChatGPT cites an average of 15.4 sources per answer, versus 11.4 for AI Mode, 9.2 for AI Overviews, and 3.3 for Gemini. (Semrush 2026 AI Visibility Index, June 2026)

Across 126 million US prompts, only 30% of the brands Gemini mentions are also cited sources, versus 64% on AI Overviews. Counts depend on method, and Muck Rack's in-text count above puts ChatGPT closer to 5. Either way, Gemini gives you very few seats.

On Google AI Overviews in Norway, 73.0% of Reddit citations pointed to Reddit's machine-translated versions of threads. (Peec AI, June 2026)

Poland (71.7%), Sweden (71.6%), and Germany (52.2%) showed the same pattern across 64.77 million Reddit citations from March to June 2026, while English-speaking markets stayed under 1%. ChatGPT's share of translated Reddit citations fell from 6.14% in April to 0.66% in May. Machine translation is quietly letting English-language conversations win non-English answers.

About 57% of AI citations go to company-owned websites. (Profound, July 2026)

Across 11.84 billion citations on eight platforms, ChatGPT sits lowest at 47% and Gemini highest at 69%. This looks like it contradicts Muck Rack, but the definitions differ: Profound counts any company's own site as a brand citation, while Muck Rack counts another company's content as earned media unless that company is the subject of the prompt. Your own site and what others say about you both feed the answer, so invest in both.

75% of pages cited by LLMs were updated in the past year, but only 42% were first published in the past year. (Seer Interactive, July 2026)

Across 47,097 citations, Gemini leaned hardest on freshness (78% updated within a year), then ChatGPT (73%) and Perplexity (65%). The sample covers four brands, so treat it as directional. You don't need new pages. You need maintained ones.

On ecommerce prompts, 37% of the URLs Gemini meaningfully cited each week were new, versus 3% of the brands it mentioned. (BrightEdge, July 2026)

ChatGPT showed the same split (15% of URLs versus 5% of brands), measured weekly on ChatGPT, Gemini, and AI Overviews from May to July 2026. BrightEdge also reported YouTube's share of ecommerce AI Overview citations climbing from about 31% to about 65% over 11 weeks, though its own share figures don't fully reconcile. The evidence turns over. The winners don't.

On location-based questions, MapQuest earned 5.8 AI citations for every one Wikipedia earned, and 80% of all citations pointed to sources a brand can influence. (Yext, July 2026)

Across 155.5 million citations for 1,623 brands in Q1 2026, the influenceable share (a brand's own website or business listings) ranged from 85% for OpenAI models to 69% for Anthropic's, and reviews and social made up 19.5% of Anthropic's citations versus 2.6% of OpenAI's. Yext sells listings management, so weigh it accordingly. For local search, your website and listings are the answer.

ChatGPT now issues an average of 7.61 fan-out queries per prompt, up from 2.17 in 2025. (Nectiv, August 2026)

About 64% of those sub-queries use a site: operator to go straight to specific domains. The data comes from roughly 4,000 prompts run through the OpenAI API, so it won't match the consumer app exactly; AirOps saw most consumer queries spawn two sub-queries earlier in 2026. Either way, the trend is more sub-queries and more direct lookups of official sources.

Reddit's share of ChatGPT citations fell 95% in a single week in August 2026. (Qwairy, August 2026)

Reddit dropped from 2.05% of ChatGPT Search citations on August 1 to 0.07% on August 14, while searches using site: jumped from 0.03% to 9.7%. Promptwatch and BrightEdge saw the same collapse, though the cause is disputed. Months earlier, Tinuiti and Profound had Reddit nearly doubling its share of commercial citations. Platforms can rewrite the source mix overnight, so never let one third-party site carry your visibility.

Creator and social platforms' share of AI citations grew 140% between August 2025 and June 2026, while owned brand and product pages' share fell 10%. (AirOps, August 2026)

Across about 3.5 billion tracked citations, YouTube's share rose 158% to 4.6% of all AI citations. The split by engine is stark: creator sources made up only 0.9% of ChatGPT's citations but about 1 in 9.5 of Google AI Overviews' citations. If your brand isn't on YouTube, you're missing Google's favorite creator source.

50% of the YouTube URLs Google AI Overviews cites carry a timestamp, versus 0.6% for ChatGPT. (BrightEdge, August 2026)

Shorts made up 17.5% of AI Overviews' YouTube citations versus 1.1% for ChatGPT, across nine industries from May to August 2026. BrightEdge puts it plainly: Google is navigating inside videos while ChatGPT is selecting between them. Chapter your videos and write transcripts that answer specific questions.

Google's AI Mode cites google.com for 44.9% of its sources, versus 9.9% in AI Overviews. (BrightEdge, September 2026)

For consumer product and finance queries, AI Mode builds answers from Google's own shopping grids, rate tables, and stock data, and the gap widened from January to August 2026. It's a narrow sample, so don't generalize it to every query. If you sell products, though, your product feed is now one of your most important citation sources.

GPT-5.6 "Luna" used a site: search in 76% of its fan-out queries, up from 22% for GPT-5.5. (Seer Interactive, September 2026)

Across 346 matched prompts, Luna gathered 43% fewer sources and gave 12.9% fewer inline citations, and on non-brand prompts its site: use jumped from 20.6% to 77.3%. ChatGPT increasingly decides which domains to check before it searches. Being the known, trusted domain for your topic matters more than ever.

When AI engines were asked to evaluate Cloud 100 companies on pricing, the company's own pricing page was cited first in only 12% of responses. (Profound, September 2026)

Owned pricing pages appeared somewhere in 46% of about 7,600 responses, while Vendr (18.7% of runs), Reddit (18.6%), and G2 (15.9%) were the most-cited third parties. Only 57 of 77 public pricing pages were fully readable by bots. Your pricing page is one of your most important GEO assets, and plenty of them can't even be read.

Community sources like Reddit and YouTube make up about 14% of AI citations for everyday products but only 6% to 9% for most health and finance topics. (Semrush, September 2026)

For health insurance and prescription drugs, more than 11% of citations come from government health sources like healthcare.gov, CMS, FDA, and NIH. Brand-owned sites got 20.6% of citations for credit cards but only 0.6% for legal services. In YMYL categories, AI goes back to the authorities, so your credibility signals matter more than community buzz.

↑ Jump to Top

Brand visibility and volatility

AI answers are probabilistic, fragmented across engines, and concentrated on a few winners, so single-prompt rank tracking tells you almost nothing.

Ask an AI the same brand recommendation question twice and there's less than a 1 in 100 chance you'll get the same list of brands. (SparkToro, January 2026)

The odds of the same list in the same order are about 1 in 1,000, based on 2,961 runs of 12 prompts across ChatGPT, Claude, and Google's AI. How often a brand appears is meaningful, though: City of Hope showed up in 69 of 71 ChatGPT answers. AirOps found the same instability in citations, with only 2.3% of page-query pairs cited in all three identical runs. Stop reporting AI rankings and track share of appearances across many runs.

Any two AI engines share only 16% to 59% of their top cited sources but 36% to 55% of their top recommended brands. (BrightEdge, April 2026)

The engines disagree about the evidence but agree more about the winners. That's good news for strong brands, because enough authority gets you recommended by different engines through different paths. The brand is the constant. The citations are the variable.

Week to week, 74% of the domains ChatGPT Search cites are new, compared with 56% in AI Mode and 5% in AI Overviews. (SISTRIX, May 2026)

In 53% of AI Overview prompts, not one source changed over 17 weeks, and AI Mode kept a stable core of one to five domains in 86.5% of prompts. For ChatGPT, the median prompt across 82,619 prompts in six countries had no domain present in all 17 weeks. AI Overviews reward holding position. ChatGPT rewards showing up everywhere, constantly.

91% of AI citations appear in only one engine, and just 2.37% of cited URLs show up across ChatGPT, Perplexity, and AI Overviews for the same prompt. (Growth Memo, May 2026)

The study covers 3.7 million citations and 20,000 prompts queried at the same minute twice a day, with a sample that skews European and toward fintech, travel, and SaaS. Winning one engine tells you almost nothing about the others. Measure each surface on its own terms.

Brands ranked #1 in frequently cited third-party listicles saw an estimated 16.5-point visibility lift in B2B SaaS. (Peec AI, May 2026)

The lift was 13.4 points in Emerging MarTech, based on about 200,000 responses across eight AI engines, and extra placements stopped adding much after a few. Peec says these are associations, not proof of cause. Still, earned placement in the lists AI already reads is one of the few levers with a measurable payoff.

61.7% of AI appearances are ghost citations, where a brand's site is cited but the brand isn't named in the answer. (Semrush, June 2026)

Another 25.1% are mentions without a citation, and only 13.2% are both. ChatGPT cites 87% of the time but names the brand in just 20.7% of answers, while Gemini names brands 83.7% of the time but cites only 21.4%. A citation count alone overstates your visibility, so track mentions and citations separately.

In News and Media, the three most visible brands hold 82.9% of all AI visibility. (Semrush 2026 AI Visibility Index, June 2026)

The top three hold 76.9% in Consumer Electronics and 41.4% in Finance, across 126 million US prompts. Only 36 brands, names like YouTube, Amazon, and Walmart, stayed in the top 100 on all four platforms every month. AI search is a winner-take-most market. Outside the top tier of your category, you're close to invisible.

After ChatGPT switched from GPT-5 mini to GPT-5.5, 47% of citations in German-language answers shifted within 48 hours. (SISTRIX, June 2026)

A normal weekday shifts 1% to 2%, based on 38 daily samples of 100,000 German-language responses in May 2026. Comparing 800,000 responses from the four days before and after the May 23 switch, Reddit gained 59% and German publishers like welt.de gained 99%, while aggregators like Tripadvisor (down 51% to 53%) and Expedia (down 60%) lost ground. ChatGPT has core updates now, and they don't come with a citation changelog.

On purchase prompts, only 40% of recommended brands overlap between two runs of the same prompt. (Conductor, June 2026)

On comparison prompts, the same brand led 91% of the time, across 14,000 API calls to four LLMs. The recommendation set wobbles, but the leader of a head-to-head holds. Win the comparison and the consistency follows.

Only 15.2% of 1,094 ChatGPT topic categories have a clear brand owner. (Semrush, July 2026)

31.2% had an emerging leader and 53.7% were unsettled, across 50,000-plus brands tracked monthly from January to June 2026. Once a brand owned a category, it held first place in 90.4% of month-over-month comparisons. Most categories are still up for grabs, and the winners tend to stay winners.

The median brand sees an 8-point daily visibility gap between its best and worst Google AI surface. (Profound, July 2026)

More than one in three brands see the gap exceed 10 points across Gemini, AI Overviews, and AI Mode, and for the same prompt any two of those models share only about a third of the companies they name. The visibility figures come from 15,155 brand-category pairs tracked daily in May 2026. Google AI isn't one surface, so track all three.

Claude and Claude Code name only 1 in 5 of the same brands. (Profound, August 2026)

Claude searched the web in 93% of responses and Claude Code in 13%, and nearly 75% of the Claude Code agent's page visits went to documentation, informational, or pricing pages. Coding agents are their own visibility surface. If you sell to developers, your docs are your storefront.

Answer engines make inaccurate claims about brands 4.3% to 6.8% of the time. (Profound, August 2026)

Pricing and billing make up 12% of the 158,000 claims checked but 24% of the inaccurate ones, and 54% of brands had at least one wrong claim that cited their own content. BrightEdge also found AI Overviews 44% more likely than ChatGPT to criticize brands. Audit what AI says about you, then fix the source, starting with your pricing pages.

After ChatGPT's June 2026 update, citations per answer fell 10.4% and unique cited domains dropped in 56 of 58 industries. (Profound, August 2026)

Nine of ten leading brands in the median industry held their position anyway. In Profound's words, the update rebuilt the source layer, not the leaderboard. Strong brands survive model changes, while the pages that cite them get swapped out.

Forcing ChatGPT to search changed brand visibility by 20 points on average for prompts that never search on their own. (Graphite, August 2026)

Left alone, 58% of logged-out ChatGPT responses to entity-comparison prompts searched the web in early August 2026. Graphite found several prompt-tracking datasets with citations in nearly every response, which suggests they force search. If your tracking tool doesn't behave like a real user, it isn't measuring what real users see.

AI platforms declined to answer branded prompts about Seer 31.2% of the time, but were 95.9% accurate when they did answer. (Seer Interactive, August 2026)

The single-brand study tracked 1,562 branded prompts across six platforms from May to June 2026, and comparison prompts were declined 80% of the time. For Seer, silence was a much bigger problem than hallucination, with just 2.8% of responses inaccurate. Give the engines something to say.

ChatGPT named 27.9% fewer B2B tech brands per answer in September 2026 than in March. (BrightEdge, September 2026)

Answers dropped from 5.36 brands to 3.87, with the steepest cut at the informational top of the funnel (down 32%), even as citations on those answers rose. The shortlist is shrinking right where buyers start. Fewer seats means a harder fight for each one.

Answers to the same prompt for different personas share only about 25% of their brand mentions. (Profound, September 2026)

Repeat answers for the same persona share about 40%. Income drove the clearest split: T-shirt brands recommended to low-income personas cost $8 to $24, versus $140 to $1,650 for high-income ones. There is no single AI answer for your category, so simulate the personas who actually buy from you.

The median page cited by AI engines loses half its citation share within 11 days of peaking. (Profound, September 2026)

78% of cited pages fall to half their peak within two weeks, based on 883,000 pages tracked daily across seven AI surfaces for a year. Pages cited by five or more engines were 5.5 times as likely as single-engine pages to still hold half their peak at day 56. Scrunch measured a longer half-life of about 4.5 weeks with a different method, with ChatGPT turning over fastest. Either way, AI visibility decays fast. Run it as a program, not a project.

Tracking 50 prompts once a day left about 6.1 points of uncertainty in a monthly visibility score, versus 13.2 points for 10 prompts run five times a day. (Peec AI, September 2026)

Both scenarios use the same 1,500 runs a month, modeled from variance in about 120 million chats, and daily swings under 12.4 points could be chance. Profound reached a similar conclusion: one run a day lands within about 2 points of ten runs a day. Breadth beats repetition, so build your tracking around prompt coverage.

↑ Jump to Top

Consumer behavior and trust

Most people read AI answers and trust them more the more they use them, but most still verify, and few click through.

Only about a third of consumers think generative AI chatbots are as effective as search engines for learning new information. (Gartner, January 2026)

82% have noticed AI Overviews, yet more than two-thirds keep scrolling past them, and 31% say AI summaries make them spend more time searching, not less. The samples are small and were fielded in mid-2025, so read this as direction. People are adding AI to their research, not handing their research over to it.

After an AI tool mentions a brand, 40% of US consumers search Google for it, 36% use Google to compare it, and only 8% ignore it. (Similarweb, March 2026)

34% ask the AI a follow-up and 28% go straight to the brand's site, based on a January 2026 survey. Three of the top four actions leave no trace back to the AI mention. The AI conversation starts the journey, and Google often finishes it.

35% of US consumers rated AI tools most useful at the product-discovery stage, versus 13.6% for search engines. (Similarweb, March 2026)

AI kept its edge at evaluation (32.9% versus 15%) before the gap nearly closed at final purchase (24.3% versus 22.1%). AI owns the top of the funnel. Search still closes.

51% of Americans aged 14 to 29 use AI daily or weekly, flat from a year earlier, while their skepticism climbs. (Gallup, April 2026)

Nearly four in ten young people expect AI to hurt their ability to find accurate information. The generation everyone assumes is all-in on AI is split on it. Don't build your plan on a stereotype.

54% of US consumers say they had to double-check all of the information generative AI tools gave them. (Gartner, May 2026)

62% say AI-provided information ended up wasting their time, and 72% say generative AI shows up in their apps "whether I asked for it or not." Call it the verification tax. The brand that shows up as the corroborating source when people double-check wins twice.

60% of US adults say they ever read the AI summaries at the top of search results. (Pew Research Center, June 2026)

Reading the summary isn't the same as trusting it, but it tells you where attention goes. For a majority of searchers, the AI Overview is now part of how they learn about your category and your brand. Make sure it's saying the right things.

76% of Americans who don't use chatbots say distrust of their accuracy is a reason why. (Pew Research Center, June 2026)

45% call it a major reason. Only lack of interest (83%) and privacy concerns (79%) are cited more often, and about two-thirds of non-users say they're unlikely to start in the next year. About half of US adults don't use chatbots, and most of them aren't planning to. That's one more reason classic search still earns its budget.

Only 4% of news consumers say they always or often click through from AI answers to the original sources, versus 19% from search. (Reuters Institute Digital News Report 2026, June 2026)

The 4% is a share of all respondents across 27 markets. Among people who actually use chatbots for news, 42% say they always or often click through, close to the 44% of search users who do (Reuters Institute). In the US, Gallup found just 7% rely on AI tools for news a great deal or a fair amount. The people using AI do click. There just aren't many of them yet.

54% of US consumers say AI is more helpful than traditional search, down from 82% a year earlier. (Fractl, June 2026)

The share rating AI less helpful rose from 3% to 17%, and for trusted product recommendations, consumers picked Google results (39%) over AI tools. Usage is climbing while the novelty wears off. AI search now has to earn trust the old-fashioned way, by being right.

People who got a brand recommendation from an AI platform, and weren't already using that brand, became 182% more likely to search it on Google within a week. (Scrunch, June 2026)

They were also 117% more likely to visit the brand's site and 185% more likely to view it on a retailer's product page, based on an opt-in panel linking AI conversations to web activity from February to May 2026. A recommendation did roughly double the funnel-moving work of a plain mention. Some of your branded search growth is AI visibility you're not getting credit for.

The share of US English-language ChatGPT conversations with commercial intent rose from 13.9% to 19.2% between June 2025 and June 2026. (Profound, August 2026)

Across 7.5 million conversations, most of the share gain landed in the final quarter, and commercial conversations per active user per week rose 41%. Commercial share ran from 59.1% in Automotive and 56.1% in Travel and Hospitality down to 6.2% in Education. ChatGPT is becoming a shopping surface, whether or not the checkout happens there.

45% of daily AI users in the US trust AI "completely or a lot" to give accurate information, versus 8% of people who have never used it. (Gallup, September 2026)

Trust falls to 26% among weekly users and 15% among less-than-weekly users, and the global median across 37 countries is 36%. Trust follows usage. As daily use grows, AI answers carry more weight with the people who see them most.

57.5% of US AI users have decided not to buy something because of information from an AI chatbot. (Semrush, September 2026)

The July 2026 survey of 2,338 US adults, run by Semrush's Exploding Topics team, found the same share of AI users (57.51%) have bought something based on a chatbot recommendation. 74.15% of all respondents would be at least somewhat less likely to buy if a chatbot flagged mixed or negative reviews. AI doesn't just recommend you. It can talk people out of you.

91.4% of 80,890 real fashion and beauty prompts to ChatGPT, Gemini, and Perplexity name no brand. (Profound, September 2026)

A companion analysis of 66,740 banking prompts found seven in ten never mention a provider. People ask AI what to buy, not whether to buy you. The shortlist is open, so earn your way onto it.

↑ Jump to Top

AI shopping and agentic commerce

Shoppers use AI to research and shortlist, and those visits now convert better than other channels, but most people still won't let an agent complete the purchase.

Generative AI traffic to US retail sites grew 693.4% year over year during the 2025 holiday season, and it converted 31% better than other traffic. (Adobe, January 2026)

Revenue per AI visit rose 254%, and AI traffic converted 54% better on Thanksgiving and 38% better on Black Friday. A year earlier AI shoppers converted worse than everyone else, and now they convert better.

Salesforce credits AI and agents with 20% of global online holiday sales in 2025, or $262 billion. (Salesforce, January 2026)

This is a broad measure that credits retailers' own personalized recommendations and customer engagement, not just ChatGPT and Perplexity. Still, the share of traffic from third-party AI search doubled year over year, and shoppers referred by AI search converted 9x more often than social referrals. AI is already in the checkout path, even when it isn't the one pressing buy.

US ecommerce sales through AI platforms will top $20 billion in 2026 and $144 billion by 2029. (EMARKETER, January 2026)

That 2029 figure would be 8.8% of US retail ecommerce. EMARKETER notes most AI-driven transactions still close on retailer websites. Even so, it's big enough to plan around.

45% of consumers turn to AI during their buying journeys, and 41% use it to research products. (IBM and NRF, January 2026)

33% use AI to interpret reviews and 31% to hunt for deals, across 18,000-plus consumers in 23 countries, yet 72% still shop in stores. AI is shaping the decision before the shopping starts, online or off.

ChatGPT's top recommended product appeared in Google Shopping's first three results 75% of the time. (Semrush, January 2026)

Semrush decoded ChatGPT's hidden shopping searches across 100 product prompts, a small sample. A separate analysis of Peec AI data found 83% of 43,000-plus ChatGPT carousel products strongly matched products in Google Shopping's top 40 organic results, versus 11% for Bing. As of early 2026, your Google Shopping rankings were feeding ChatGPT's shelf.

Organic search sent about 119 times more visits than AI referrals to 19 major retailers during the 2025 holiday season. (Semrush, January 2026)

That's roughly 1.055 billion organic visits versus 8.87 million AI referrals from November 1 to January 9, with AI at about 0.15% of traffic, though AI referrals jumped 213% on Cyber Monday versus early November. Hold that next to Adobe's growth rates above. AI shopping traffic is growing fast from a tiny base.

Amazon says 300 million-plus customers used its Rufus shopping assistant in 2025, helping deliver nearly $12 billion in incremental annualized sales. (Amazon, February 2026)

By Q2 2026, Amazon had folded Rufus into Alexa for Shopping, with active users close to doubling and interactions up more than 5x year over year. The biggest store on the internet now runs its own answer engine. If you sell on Amazon, that assistant is part of your AI search plan.

OpenAI pulled back ChatGPT Instant Checkout in March 2026 with somewhere between a dozen and 30 Shopify merchants live. (Forrester, March 2026)

Checkout is moving to merchant apps inside ChatGPT or back to merchant websites, which makes the answer engine a referral source again. Forrester's own survey found completing a purchase inside an answer engine was consumers' least-adopted use case. Agentic checkout is coming, just slower than the hype.

38% of European consumers use AI tools to research products or decide what to buy, but 47% are uncomfortable letting AI manage recurring purchases. (McKinsey, March 2026; McKinsey, April 2026)

McKinsey found trust declines steadily as AI gets closer to execution: prefilled baskets, checkout, and auto-reordering. 56% are comfortable with AI suggesting options as long as a human makes the final call. For now, people use AI to browse, not to buy.

ChatGPT's shopping widget appeared in 87% of responses to shopping prompts in February 2026, up from 8.3% in October 2025. (Evertune, March 2026)

That's a fixed basket of 200-plus "best products" and "best deals" prompts, and the widget topped 90% in 8 of 11 categories. Across all tracked prompts, not just shopping ones, Profound found Shopping cards on 8.82% of submissions. When the prompt is a shopping question, ChatGPT now answers with products.

Walmart holds 8.78% of first-position buy links on ChatGPT Shopping product cards, and Target leads total buy-link presence at 7.16%. (Profound, March 2026)

Across 22.5 million buy offers in March 2026, the top 20 merchants held about 40% of all offers. In separate carousel tracking that month, about 95% of product titles appeared in fewer than 30% of runs for the same prompt. The big-box retailers lead on the buy button. The carousel underneath them reshuffles constantly.

AI traffic to US retail sites converted 42% better than non-AI traffic in March 2026, after converting 38% worse a year earlier. (Adobe, April 2026)

39% of US consumers have used AI assistants for online shopping, and during Prime Day 2026 AI traffic converted 40% better than non-AI channels such as paid search, email, and social. Adobe also audited retailer pages and found product pages average just 66% on machine readability. The AI shoppers are arriving. Most product pages aren't built for the machines sending them.

74% of AI Mode users' final shortlists came straight from the AI's answer, with no outside check. (Growth Memo, April 2026)

In a usability study of 48 US participants making 185 high-stakes purchase decisions (TVs, laptops, appliances, and car insurance), AI Mode users rarely left the answer. It's a small qualitative sample, but it shows the stakes. When AI Mode builds the shortlist, you're either in the answer or out of the deal.

Willingness to let AI make a purchase decision tops out at 11%, even for low-stakes categories like household supplies. (Gartner, May 2026)

31% would let AI narrow their choices for household supplies, and 28% for personal electronics. Menlo Ventures found 37% of AI-using shoppers made a final decision based on an AI recommendation, but that measures past behavior among AI shoppers, not willingness across all consumers. Either way, the shortlist is where AI holds its power. Get on it.

Merchants in the US and UK say only 3% of transactions involve AI agents today, and 24% of consumers say they'll never delegate a purchase to AI. (Checkout.com, June 2026)

A third of consumers expect at least 10% of their purchases to be AI-driven within a year, and their top requirements are spending caps (30%), instant revocation (29%), and easy cancellation (28%). 89% of merchants say they're preparing. The gap between expectation and trust is where agentic commerce gets decided.

Alibaba is ChatGPT's most-cited retail source at 4.2 million citations, almost twice Amazon's count. (Semrush, June 2026)

Amazon, meanwhile, accounts for 42% of all retail brand mentions in AI Overviews and is the most-cited retail source on AI Mode and AI Overviews. Of 124 retail brands tracked, 39 appeared on just one platform. Retail visibility in AI depends on which engine you're looking at.

Shopify says AI-driven traffic and orders to its merchants' stores tripled year over year in Q2 2026. (TechCrunch, August 2026)

Half of AI-referred sessions land directly on a product page, 2.5x the rate for traditional search. Shopify President Harley Finkelstein also said traditional search sessions are up 1.3x over two years and still make up about a third of storefront sessions. AI search is adding shoppers, not replacing Google, and your product pages are the new landing pages.

When the request to AI coding agents moved from a general recommendation to a concrete setup task, they changed their first-choice company in 232 of 449 matched comparisons. (AirOps, August 2026)

Across 1,350 trials of Claude Code, Codex, and Cursor on 50 enterprise software companies, GitHub's win rate as an alternative fell from 95.2% to 36.5%, while MongoDB's rose from 6.1% to 32.3%. The three agents picked the same winner only 57.3% of the time on general recommendations. What an agent recommends and what it prefers to implement are different contests.

The day after GPT-5.6 launched, feed-integrated retrieval in ChatGPT Shopping jumped from 8.26% to 61.54% of retrievals. (Profound, September 2026)

The daily split came from 1.76 million Shopping prompt runs in July 2026. Comparing July 7 to 9 with July 10 to 12, the top 10 merchants' share rose from 22.5% to 41.8% and unique merchants fell from 13,524 to 10,607. Shopify accounts for about 35% of feed-integrated retrieval. If your products aren't in a feed ChatGPT can read, you're fighting for a shrinking slice of the shelf.

US desktop shopper journeys that used both AI and search converted at 23.0%, versus 17.3% for search alone and 12.5% for AI alone. (Similarweb, September 2026)

AI-and-search journeys touched 14.9 sites, compared with 10.1 for search alone and 7.3 for AI alone, in June 2026. In 89% of shopper journeys that involve AI, search is there too. The winning journey isn't AI or search. It's both.

In mid-August 2026, a ChatGPT shopping experiment running on OpenAI's own product index was active in 8% of chats. (Peec AI, September 2026)

A Peec researcher found the test, named "prefer-index-over-serp-v3," while inspecting the data ChatGPT streams to the browser, with OpenAI testing it on traffic slices from single digits up to 20%. It's one researcher's inspection, not a sampled study. If OpenAI moves shopping off Google's results and onto its own index, the rules for ChatGPT product visibility change again.

↑ Jump to Top

Marketer adoption, budgets, and the GEO industry

Budgets for AI search optimization nearly doubled in a year, B2B buyers now start in chatbots, and measurement is the weak link.

94% of enterprises plan to increase AEO/GEO investment in 2026, after spending an average of 12% of digital marketing budgets on it in 2025. (Conductor, January 2026)

The survey covers 250-plus senior executives and digital decision-makers at enterprise organizations, and AEO/GEO ranked as their top marketing priority for 2026. Conductor sells AEO software, so weigh it as vendor research. The top challenges were trusting the quality of AI visibility data and scaling AI-optimized content, which is exactly where most teams are stuck.

94% of business buyers use AI during their buying process. (Forrester, January 2026)

When the purchase includes generative AI features, the buying group doubles from 7 people to 14. Forrester warns that answer engines can breed mistrust with incomplete or unreliable information, so buyers check AI output with peers and analysts. You have to show up in the AI answer and in the sources buyers use to check it.

51% of B2B software buyers now start their research with an AI chatbot more often than with Google, up from 29% a year earlier. (G2, April 2026)

69% chose a different vendor than planned because of a chatbot's recommendation, and 33% bought from a vendor they hadn't heard of before. 85% think more highly of vendors that AI includes in its answers. In B2B, the AI shortlist is the shortlist.

Peec AI passed $10 million in ARR 16 months after launch, with 2,500-plus customers. (Peec AI, May 2026)

Its ARR doubled from $5 million in five months. Six days later, Sitecore acquired Scrunch, an AI visibility platform with 500-plus customers, for a reported $225 million. AI visibility tooling went from experiment to M&A in about 18 months.

Only 22% of marketers say their SEO and AI search efforts are fully integrated, and just 9% can measure all the metrics that matter. (Semrush, June 2026)

40% still check AI visibility by typing prompts into AI tools by hand, and only 14% plan to invest in measurement versus 38% in AI search optimization. Fully integrated teams were far more likely to report AI-driven traffic or leads (81% versus 36% for fully separate teams). Integration is the multiplier. Build one team, not two.

Active ChatGPT advertisers jumped 46% in one week, from 2,585 to 3,770, in the week ending June 8, 2026. (Similarweb, June 2026)

Similarweb tracked 5,171 unique advertisers from March 30, with the top five capturing nearly 17% of global impressions and Monday.com alone 5.05%. The US held 71.6% of impressions. The early movers are already in.

83% of ChatGPT prompts that triggered an ad would never have triggered a Google Shopping ad. (Similarweb, June 2026)

Google AI Mode ads, by contrast, are keyword-matched and fire on the first turn 98.5% of the time. ChatGPT's ad inventory lives in conversations your keyword lists don't cover. Paid AI search needs its own playbook, not a copy of your Shopping campaigns.

Marketers put an average of 24% of their search or content budgets toward AI search visibility, yet 81% still call the work SEO. (Fractl via Search Engine Land, July 2026)

82% have allocated at least some budget, and C-suite respondents use "GEO" and "AEO" about three times as often as practitioners. 66% of marketers have used AI tools to research marketing vendors, so your agency shortlist is being built the same way. The label fight doesn't matter. The work does.

82% of B2B software buyers got software recommendations from an AI chatbot in the past 24 months, but only 9% are comfortable letting AI agents buy within approved guardrails. (G2, July 2026)

Only 2% would let agents buy without pre-approval, even though 61% use or plan to use agents somewhere in the process. Evaluation now takes up 40% of the buying journey. AI makes software easier to find and harder to buy, so your visibility has to hold up through a longer evaluation.

22% of CMOs now start vendor research in an AI answer engine, ahead of the 16% who start with Google or another search engine. (Profound, July 2026)

Only peer recommendations (49%) ranked higher among the 100 CMOs interviewed, and 90% use answer engines daily. One in five of the people who sign off on your deal now start vendor research in AI.

92% of US business buyers who use AI at work say it has shaped their vendor shortlist, and 45% say it did so significantly. (Semrush, July 2026)

After an AI mention, 71% visit the vendor's website and 63% search for the company on Google, based on 519 AI-using respondents surveyed in March and April 2026. Only 7% notice a vendor in an AI response because they already recognize the name. AI is introducing buyers to vendors they didn't know, so brand awareness is no longer the price of a shot at the deal.

ChatGPT's ads business hit a $1 billion annualized run rate in under 200 days. (Digiday, August 2026)

Ads launched on February 9, 2026 on the Free and Go tiers and opened to 31 European markets in August with self-serve Ads Manager. A run rate isn't bookings, and Digiday estimates far less was actually booked through August. Either way, AI search now has a paid layer, and budgets will follow it.

In 96.37% of ChatGPT ad placements on commercial prompts, the advertiser wasn't among the answer's cited sources. (SE Ranking, August 2026)

Across 50,006 US commercial prompts in 20 niches, ChatGPT showed ads on 25.94% of them. Healthcare prompts showed ads 28.69% of the time in ChatGPT versus 2.64% in AI Mode, and about 1 in 7 ads had no topical connection to the prompt. Buying the ad doesn't buy you the answer.

69% of search professionals now have budget allocated for AI search optimization, up from 38% in 2025. (SEOFOMO, September 2026)

92% track AI visibility and citations, and 98% have had a client or boss ask about it. The payoff is murky: 46% attribute only 0% to 5% of site revenue to AI search, and 35% can't tell what it contributes. It's a small, self-selected sample of 171, but the pattern matches everything else here. Spending is ahead of measurement.

Profound raised a $180 million Series D at a $1.8 billion valuation, less than seven months after a $96 million Series C at $1 billion. (Profound, September 2026)

Profound reports more than 1,000 enterprise customers, and it told TechCrunch its revenue tripled in six months. Its valuation rose 80% since its February Series C. Brands will pay real money to see how AI talks about them.

ChatGPT showed ads on 26% of US desktop responses in June 2026, up from 14% in May. (Similarweb, September 2026)

66.3% of ChatGPT ads appear after the second prompt or later, 65.3% of sessions continue after an ad shows up, and ad click-through was 0.50% in June. Trackers measure ad load differently, and Evertune put it at about 13% of responses in late May and early June. Either way, the paid layer is real, and Similarweb's numbers show it nearly doubling in a single month.

ChatGPT ads went from nearly 0% of tracked fintech answers in April to 53.1% by July 19, 2026. (AirOps, September 2026)

Across 2.8 million answers for 131 fintech brands, the brand named in a prompt got just 3.1% of the ad placements on that prompt, so another company usually took the only paid slot. In 92.7% of ads with organic brand context, the advertiser was absent from the answer itself. Your brand name is now something competitors can buy against in ChatGPT.

86.6% of marketing leaders are prioritizing AI search, but only 23.2% feel confident measuring its impact. (AirOps, September 2026)

The Marketing Leaders Reality Index, a survey of 307 leaders fielded in July and August 2026, also found 75.4% had higher pipeline or revenue targets while only 43.0% got any budget increase. Priorities are up, budgets aren't keeping pace, and measurement is the bottleneck.

↑ Jump to Top

Publishers, crawlers, and the open web

Bots now make up more than half of internet traffic, Google referrals to publishers fell by a third or more, and AI apps send back almost nothing. The crawl-for-clicks bargain is breaking.

Google organic search traffic to more than 2,500 news sites fell 33% globally and 38% in the US between November 2024 and November 2025. (Reuters Institute, January 2026)

The Chartbeat data behind it covers 2,576 sites, and the publishers surveyed expect search traffic to fall another 43% over three years. The report notes it's unclear how much of the drop comes from AI Overviews, since hard news queries have been largely exempted from them. Chartbeat's latest release, in September, still shows Google referrals down 36% year over year.

Google sends news publishers 500 times as many referrals as ChatGPT from Search alone, and 1,300 times as many including Discover. (Reuters Institute, January 2026)

ChatGPT referrals are growing fast but remain "little more than a rounding error," and only 20% of publishers expect substantial AI licensing revenue. The traffic AI features take out of Google isn't coming back through chatbots. Not yet, and maybe not ever.

Googlebot saw about 1.7x as many unique URLs as ClaudeBot, 1.8x as many as GPTBot, and about 167x as many as PerplexityBot over two months. (Cloudflare, January 2026)

Because one crawler feeds both Search and Google's AI, publishers can't block Google's AI use without losing search visibility. Cloudflare argues that's why Google-Extended-style opt-outs don't level the field. It's the Google advantage in crawl form.

By Q4 2025 there was one AI bot visit for every 31 human visits to publisher sites, up from 1 in 200 in Q1 2025. (TollBit, February 2026)

Click-through from AI apps fell from 0.8% to 0.27% for sites without AI deals, and from 8.8% to 1.33% for sites with licensing deals. A licensing deal doesn't protect your referral traffic. It just changes how you get paid for losing it.

Serving Markdown instead of HTML to AI bots produced about one extra median bot visit per page over three weeks, a result that wasn't statistically significant. (Profound, February 2026)

The randomized test covered 381 pages on six sites. ChatGPT-User made up about three-quarters of bot traffic and showed about a 20% lift for Markdown, also not significant. Before you rebuild your site for bots, test it. The hype is running ahead of the evidence.

Search referral traffic fell 60% for small publishers, 47% for medium publishers, and 22% for large ones over two years. (Chartbeat via Axios, March 2026)

From December 2024 to December 2025, Google Search referrals fell 34% and Discover 15%, while ChatGPT referrals grew more than 200% from a tiny base. Chatbots still account for less than 1% of publisher page-view referrals. Scale is protection. Small sites without brand demand are taking the hardest hit.

OpenAI's crawling reached about 4% of Google's total crawl volume, up from 1.38% a year earlier. (Botify, April 2026)

In 7 billion-plus log lines analyzed by Chris Long of Nectiv, OAI-SearchBot activity rose 3.5x and GPTBot 2.9x since August 2025, while ChatGPT-User fell 28% from December 2025 to mid-March 2026. OpenAI is crawling far more of the web than it did a year ago. Check your logs for these bots and make sure your server isn't turning them away.

OpenAI accounts for 96% of live AI user-agent activity, where AI fetches pages on a user's behalf. (BrightEdge, June 2026)

BrightEdge counts 88 agentic requests for every 100 organic search visits, and OpenAI also makes up 51% of training crawls and 49% of search-index crawls. BrightEdge doesn't give a sample size. When an agent visits your site right now, it's almost always ChatGPT.

28% of 137,000 domains publish an llms.txt file, but 97% of those files got zero requests in May 2026. (Ahrefs, June 2026)

96% of the requests that did arrive came from bots, and named AI bots made just 19.5% of them. Ahrefs' customers skew technical, so 28% is likely an upper bound on adoption. llms.txt is a file almost nobody reads, so spend the effort on pages bots actually crawl.

In ChatGPT Deep Research's first read of a page, pages with fewer than 20 links were about 78% real content, versus about 33% for pages with 60 or more links. (Peec AI, June 2026)

A first read returns a capped window of about 5,700 characters at the median, and every search in Peec's logs ran on Bing. Heavy navigation eats the agent's reading budget before it reaches your content. Put the answer high on the page and keep the chrome light.

More than 50% of traffic on the internet is now non-human. (Cloudflare, July 2026)

Cloudflare says agent traffic "crossed a historic threshold for the first time" in 2026, and some of the most heavily crawled categories saw human traffic fall as much as 40% in less than a year. Cloudflare cites its Radar data but doesn't define non-human traffic or show how it measured the milestone, so treat it as directional rather than precise. Either way, across the web, more traffic now comes from machines than from people. You're now building for two audiences.

52% of crawler requests were for AI training as of June 2026, up from 22% in spring 2025. (Cloudflare, July 2026)

Mixed-use crawlers made up more than 36%, and pure search crawling is a small and shrinking share, even though Google still drives about 88% of referral traffic across Cloudflare's network. Most of the crawling hitting your site is now for training, not for sending you visitors.

People Inc. reported a 22% year-over-year drop in core sessions in Q2 2026 and attributed it primarily to Google AI Overviews. (People Inc. Q2 2026 earnings release, August 2026)

Sessions fell from 2.2 billion to 1.7 billion, yet digital revenue still grew 6% because 43% of it now comes from sources not tied to sessions, with licensing up 23%. It's rare to see a public company name AI Overviews as the cause in an SEC filing. It's also a playbook: diversify revenue away from the click.

Combined visits to the 50 biggest US news websites fell 37% in two years, from 4.9 billion to 3.1 billion a month. (Press Gazette, August 2026)

That's July 2024 to July 2026. Organic search referrals across the top 50 were down 42% versus two years earlier, and generative AI accounted for just 0.4% of their traffic. Even the biggest brands in news aren't big enough to be immune.

In Europe, it took 179 AI scrapes to send one human visitor back to a publisher in H1 2026, versus 53 in North America. (TollBit, August 2026)

The European ratio worsened from 150 to 1 in Q1 to 227 to 1 in Q2, OpenAI's ratio went from 165 to 1 to 265 to 1, and Meta's passed 1,500 to 1. AI apps sent just 0.05% of external referrals to European sites, while Google sent about 84%. The crawl-for-clicks bargain that built the web isn't holding for AI.

About 15% of AI scrapes in H1 2026 ignored an explicit robots.txt disallow. (TollBit, August 2026)

On European sites, ChatGPT-User had the highest bypass rate at 54% of its scrapes, followed by Bytespider (48%) and PerplexityBot (42%). 60% of European sites disallow GPTBot, versus 41% in North America. Robots.txt is a request, not a wall.

Google AI Mode produced 54.3% of the AI citations for publishing brands, more than ChatGPT (19.4%) and Perplexity (13.1%) combined. (Conductor, August 2026)

Across 672 publishing domains from January to May 2026, NerdWallet (7.5%), Bankrate (6.5%), and Forbes (5.7%) led citation share. In a separate May 20 to June 20, 2026 snapshot of 16.1 million US searches, AI Overviews appeared on 41.3% of publishing searches, and 57.8% of their citations came from pages outside Google's organic top 10. For publishers, Google's AI surfaces hand out the most citations, so tracking that stops at ChatGPT misses most of them.

Fewer than 1% of Cloudflare sites block search bots, while 17% block AI training crawlers. (Cloudflare, September 2026)

Site owners want to be found but not trained on. Cloudflare's new "Accountable" designation for mixed-use crawlers, which it says Apple, Google, and Microsoft qualify for, is an attempt to give them both. Set your crawler policy on purpose, because blocking the wrong bot can take you out of the answers entirely.

↑ Jump to Top

What the numbers add up to

Five patterns run through the 2026 data.

↑ Jump to Top