Drawn from Google Search Console data for 166 brands, roughly 9.5 billion impressions and 105 million clicks a month over ~16 months.
Findings at a Glance
| Finding 1 | The −0.29 link between a site’s branded search demand and how well it ranks for the plain keyword falls once other factors such as authority are accounted for. |
|---|---|
| Finding 2 | The same positions earned fewer clicks in 2026 than they did in 2025; position three fell by more than 50%. |
| Finding 3 | Moving from position four to three is worth +18 clicks per 1,000 impressions; from position eight to seven, just +1. |
| Finding 4 | Earning more clicks than your position normally brings comes before a rank gain, an effect 2x stronger for unbranded queries. |
| Finding 5 | On weekends, B2B sites rank about 0.25 of a position worse while consumer rankings hold, and do only 21% of their searching against 28%. |
| Finding 6 | Bigger sites rank better by a wide margin: average unbranded position runs from 17 to 10 from the smallest sites to the largest. |
| Finding 7 | Mobile ranks a little better for the same query in 79% of sites. |
| Finding 8 | In the top three SERP positions, brand queries earn the most clicks (about 7%), and comparison queries the least (about 3.5%). |
| Finding 9 | At the same rank (positions 4 to 10), CTR rises from 0.5% to 1.9% as a query length grows from one word to four. |
| Finding 10 | Subdomains, deep pages, and duplicate URLs do not hurt rank, and queries where 5+ of a site’s URLs rank reach a better top spot, not worse. |
| Finding 11 | A top-three ranking is still top three a year later 76% of the time, and a top-ten ranking is 95% stable. |
| Finding 12 | Image search is 15% of impressions and under 4% of clicks. |
| Finding 13 | Search demand naming the new year overtakes the old one in January every year, so refresh year-stamped content in December. |
TL;DR:
- On unbranded queries, click-through rates are falling. Across the portfolio, the number-one position now averages about 6 clicks per hundred, and on the head terms we can track it roughly halved between 2025 and 2026.
- The one lever that comes before a ranking gain is engagement: pages that earn more clicks than their position should bring tend to climb the following week.
- Nearly all the click value is in positions one through four. Moving from eight to seven gets you almost nothing.
- Several structural worries do not hold: subdomains, deep pages, and multiple URLs ranking for one query show no penalty, and the last one runs backward.
- Update year-stamped content in December, because search demand for the new year takes over in January.
What We Wanted to Test
The claims about how Google ranking works get repeated until they become rules: brand demand lifts your rankings, subdomains hold you back, duplicate URLs drag each other down. Most started as a real observation on a handful of sites, then traveled far past the data that produced them.
iPullRank holds Google Search Console data for a large, cross-industry slice of the web, so the question was simply which of these claims survive contact with first-party data at scale, and which turn out to describe one kind of site.
Primary research question:
Which of the widely repeated claims about how ranking works hold up in first-party Search Console data across 166 brands, and which fall apart once you separate a real signal from brand size and seasonal drift?
Secondary questions:
- Does brand demand on “branded” +vanilla term move a site’s unbranded rankings for the vanilla term?
- Are click-through rates for a given position changing over time?
- Where on the page is moving up one position actually worth the effort?
- Does anything a site controls reliably come before a ranking gain?
- Do mobile and desktop rank differently for the same query, and why?
- Do subdomains, deep pages, or several URLs ranking for one query hurt?
- Once a page reaches the top, how much does its ranking move?
- At the same position, how do query intent and length change the click rate?
- When does search demand for a new year overtake the old one?
Methodology
Dataset
The data comes from Google Search Console and is read directly from the accounts of 166 brands across 272 verified properties. Every row is one query on one day: the searches a site appeared for, with clicks, impressions, click-through rate, and average position, broken out by page, device, country, and search type. Our window is March 2025 through mid-July 2026, about 16 months of continuous daily data: 151 billion impressions and 1.68 billion clicks, or roughly 9.5 billion impressions and 105 million clicks a month.
One question needed data GSC cannot see on its own. To answer whether brand demand influences unbranded rankings, we pulled competitor ranking data from Semrush to see how rival sites rank for the same terms.
Analysis Approach
The analysis focused on avoiding two common statistical traps that can produce misleading conclusions.
The first is a pattern that looks true when you line up different sites but can vanish the moment you follow one site over time, because bigger sites rank better and also attract more branded search, so those things move together without one causing the other. Every claim was checked both ways: across sites, and within a single site over its own history.
The second is drift. The portfolio’s search volume grew over the sixteen months, with more impressions and clicks as time passed, so a plain before-and-after around any event shows a rise whether the event caused it or not. That matters only for the few claims about one thing moving another over time, such as whether a ranking loss later drives more brand searches. For those, we did not trust the raw jump. We compared the real event against two controls: the same measurement after the opposite event (a ranking gain rather than a loss), and after random dates with no event at all. An effect was counted only when the real event clearly beat both, and most did not, so those claims were dropped.
Finding 1: Brand demand does not move your unbranded rankings
A popular idea holds that heavy “brand plus keyword” demand teaches Google to rank that brand for the plain keyword. In this data, it does not.
At first glance, brands with more branded-search demand do tend to rank a little better for the plain term, a correlation around −0.29 that is strong enough to clear a significance test on its own. The problem is what it travels with. Large brands attract branded searches and also carry the things that actually move rank: more authority score and more referring domains, and those track rank too. The sharper test is to stop comparing brands and follow one brand over its own history: in that view, the effect was never there. Months when a site’s branded demand jumped were not followed by gains on its unbranded terms, at any lag we checked.
None of this rules out a small true effect buried in a system with thousands of signals; a correlation that size could be real and still be swamped by everything else. What the data does rule out is brand demand as a lever you can pull, because nothing a site did to its own branded search moved its unbranded rank. Brand-building is genuinely valuable, just not through this door. Growing branded search will grow branded traffic and conversions.
Finding 2: Click-through rates are falling, and the top of the page is losing the most
The same positions earned fewer clicks in 2026 than they did in 2025.
On unbranded queries across the full portfolio, the top result earned about seven clicks per hundred impressions in 2025 and about five in 2026. The drop is steepest just below it: position two fell from about six to four, and position three from three and a half to under two, more than halving. From position four down the curve barely moved. The losses sit in the top three, which is where the clicks are.
The losses were not spread evenly down the page. Each of the top three positions shed about two clicks per hundred impressions, while position four and below gave up almost nothing.
The likely cause is the results page answering more searches on its own, through snippets, answer boxes, and AI summaries, so the same ranking turns fewer of its impressions into visits. A click curve has a shelf life: a rate measured a year ago now overstates the visits a position will bring today.
Finding 3: Almost all the click value is in the top four positions
Moving up one spot is worth far more near the top of the page than in the middle.
Climbing from position four to three adds about nineteen clicks for every thousand times a page appears. The identical one-spot move from eight to seven adds about one. The value is concentrated at the very top, so the pages worth pushing are the ones that can realistically reach the top four for a query they already rank on. Effort spent nudging a page from eight to seven, or from twelve to eleven, rarely earns back the work.
Read together with the flat curve above, this reframes where ranking effort should go. It is not “improve every position.” It is “find the pages sitting at four through seven with real demand, and get them into the top three.”
Finding 4: Better click-through rates come before better rankings
The signal appears within 1 to 2 days, and the effect is largest by 28 days.
When a page earns more clicks than its position would normally bring, its ranking tends to improve over the following week. Rankings and clicks move together for an obvious mechanical reason: a higher spot draws more clicks, so we only looked at whether a jump in clicks came before a later rise in rank, and we stripped out two things that could fake it: a ranking’s natural tendency to drift back toward its own average, and market-wide swings that lift everyone at once.
Some of the raw signal was that drift. A clear piece survives all of it. Two details make it trustworthy rather than a fluke:
- It is about twice as strong for unbranded queries as for branded ones, which is the tell that this is engagement at work rather than brand recognition.
- It helps a page climb within the top ten. It does not pull a page up out of page two, where nothing we tried moved the needle.
The click change comes before the rank change, which is strong evidence that clicks lead rank, though not the airtight proof a controlled experiment would give. The action is concrete: on pages already ranking in the top ten, rewrite titles and snippets to earn more of the clicks the position can bring, then watch whether the ranking follows. The pages that under-earn their expected click rate are the place to start.
Finding 5: B2B loses a little rank on weekends, and a lot of volume
On unbranded terms, business sites rank slightly worse on Saturdays and Sundays. Consumer sites hold steady, in rank and in volume.
I included only users with at least 5 queries on each side (3,643 users), since fewer than that gives noisy rates.
On unbranded terms, business-to-business sites sit about a quarter of a position lower on weekends than on weekdays (5.06 vs. 4.81), while consumer sites barely move (5.52 vs. 5.46). It is a small gap, but it runs about four times deeper for B2B, which fits a weekend business searcher who is more casual than the weekday one and lands on different, slightly-worse-ranking queries. The effect is modest, worth watching rather than acting on.
The bigger weekend effect is in volume, and it is by design. Split the week into weekdays and weekends, and the two audiences separate cleanly. If search were even across all seven days, the weekend would be about 29% of impressions. Business-to-business brands come in well under that, around 21%, because their searchers are at work Monday to Friday. Consumer brands sit right on the even line, around 28%.
Finding 6: Bigger sites rank better, by a wide margin
The more pages a site ranks in Google, the better its average position on unbranded terms.
Grouping sites by the number of distinct pages they rank, the fewest-page quarter averages about position 17 on unbranded terms and the most-page quarter about 10. We measure size by page count rather than by impressions on purpose: impressions are partly a result of ranking well, which would make the pattern circular, while the number of pages a site ranks is a cleaner measure of how large it is.
Ranking more is not the same as converting each position into clicks more efficiently, so we checked that too: on pages with at least a hundred clicks, we compared each page’s click rate to the rate its position usually earns. The answer depends on how you count, and the split is the point.
By the share of pages that beat their position’s expected click rate, larger sites come out ahead: about a quarter of the smallest sites’ pages over-earn, against roughly half for the biggest. Weight by traffic instead and it flips, because the largest publishers carry a heavy tail of high-impression informational queries that get answered on the results page and never earn the click. So bigger sites are not less efficient page for page. They simply also own most of the impressions that were never going to convert.
Finding 7: Mobile ranks better than desktop
A raw average shows mobile three and a half positions ahead. Part of that is a real device effect, and part is what people search on a phone.
On a blended average, mobile looks far ahead of desktop, about position 7.6 against 11.1, three and a half positions better. Almost none of that, though, comes from the device itself. Line up the exact same unbranded query on both, and the difference nearly disappears: mobile still edges ahead on most sites, roughly four in five, but only by a median of about a third of a position, with average unbranded position at 5.8 on mobile against 5.9 on desktop. What makes the blended gap look so wide is not that Google favors phones, but that people search for different things on each device, and mobile’s mix leans toward queries that already rank near the top.
The obvious guess is that mobile carries more branded search, since branded queries tend to sit at the top of the page. That is not what is happening. Mobile’s branded share of impressions is 4.9%, if anything a little below desktop’s 6.2%, and which device carries more of the branded search turns out to depend more on the industry than on the device.
Split that by industry and a clear pattern shows up. In home and local services, and in telecom, searches on a phone carry the brand name more often than searches on a desktop do. In retail and automotive, it runs the other way, with desktop searches more likely to name the brand. Everywhere else, from finance to health to software, the two devices are about even. So mobile skews more branded only where the search is local, which fits the way people reach for a phone to find a nearby business by name.
Finding 8: Whether shopping queries earn the click depends on the industry
Among queries that all rank in the top three, click rate varies with intent, but which intent wins flips from one industry to the next.
Because every query here already ranks in the top three, the differences are about intent. In retail and automotive, shopping queries earn the most clicks of any intent, around eight per hundred, ahead of brand or product searches. In finance, insurance, health, and software, the same shopping queries earn the fewest, often under two per hundred. Shopping-intent click rate at the top runs from under 0.5% in finance to more than eight percent in automotive.
The reason is what sits above the result. In commerce verticals, the top of the page still sends the click to a product or a retailer. In finance, insurance, and health, a commercial query at the top is surrounded by ads, comparison units, and answer boxes that take the click first. So ranking a commercial page in the top three does not guarantee the visit; whether it converts depends on how crowded your vertical’s results page is.
Finding 9: At the same rank, longer queries earn a higher click rate
At the same position, the click rate itself rises.
Holding position constant, among pages that all rank in positions four through ten, click rate rises with query length: from about 0.5% at one word to nearly 2% at four, then it levels off. Because the position band is fixed, length is what changes, not rank. A one-word search is broad, with someone typing a category and window-shopping the results. A four-word search is someone who already knows what they want and clicks the match. The effect holds for branded and unbranded searches alike.
Finding 10: Subdomains, deep pages, and duplicate URLs do not hurt rankings
Three long-running structural worries, tested on unbranded queries only. None of them show up.
All three are measured on unbranded queries. Branded and product-name searches naturally pull several pages from one domain to the top of the results, so leaving them in would rig the duplicate and cannibalization tests. With them removed:
- Subdomains. Pages on a subdomain average position 6.6, pages on the main site 6.3. Effectively the same.
- URL depth. Average position is flat no matter how deep a page sits in the folder structure. The homepage is the weakest page, because it competes for the hardest head terms and loses to specific inner pages built for them.
- Cannibalization. Queries where more of a site’s pages rank reach a better top position, from 4.9 with one URL to 2.3 with five or more. That is the opposite of the fear that they drag each other down.
Finding 11: Once you reach page 1, rankings barely move
Stability is real, but it is a top-of-page phenomenon, not a site-wide one.
On unbranded queries, a top-three ranking is still top three a month later about 88% of the time, and a full year later 76% of the time. Top-ten rankings barely erode, holding at 95% after a year. That calm is only near the top. Day to day, a top-three position barely twitches, while a page down around position twenty swings all over the place. The higher you rank, the less you move. Getting there is the hard part; once you are near the top, holding it mostly takes care of itself.
Finding 12: Image search is exposure, not traffic
It generates a seventh of your impressions and almost none of your clicks.
Image results are about 15% of impressions but under 4% of clicks, a click rate roughly a fifth of web search.
Finding 13: Refresh your year-stamped content in December
Demand for the new year overtakes the old one on a schedule you can plan around.
Track searches that name a year and the changeover is sharp rather than gradual. Through the autumn, searches naming the current year still dominate while the next year climbs slowly in the background. In January, the next year takes over, and by February it is most of the demand. So the window to update an article titled for a year, a “best X for [year]” or a “[year] guide”, is December, before the January flip.
The timing shifts a little by industry:
- Most categories, including finance, insurance, and software, turn over in January.
- A few, such as gaming and health publishing, lag into February.
- Evergreen reference material, like codes and standards, never really flips, because the year is not the point of the search.
What This Means for Brands
Most of the claims that compare sites, brand demand, branded share, cannibalization, URL structure, describe what strong sites look like rather than what builds them, and several run backward. The claims about a single site over time are quieter still: rankings barely move once you are near the top, and brand demand does not push them.
The one place with a real, forward-looking signal is engagement. When a page earns more clicks than its position should bring, its ranking tends to improve the following week. That is the work: write titles and snippets that beat the click rate your position would earn on its own, and spend the effort on pages already in positions one through four, where the clicks actually live. It is an engagement lever, not a branding one, and it is the only thing here that points forward instead of describing the present.
The iPullRank POV
Most SEO advice is built from between-site comparison, and this data shows how often that betrays you. Big sites rank better, attract more branded search, and hold their positions, so almost anything you measure across sites moves with almost anything else. Line the same sites up and half the usual rules reappear as descriptions of scale, not causes of it. The test that matters is whether a pattern still holds when you follow one site over its own history. Most of the claims here do not.
The one that does is engagement, and it points at how to read the rest of the study. A borrowed click curve, a blended average position, a single zero-click number: each is an average taken across devices, intents, and years that no longer behave alike. The click at position one is worth about half what it was a year ago. The same rank earns four times the clicks at four words as at one. A branded query and an unbranded one barely belong in the same table. The work is not to chase a better number against someone else’s benchmark. It is to measure your own curve, segment it by the things that actually move it, and spend effort where a click is still on offer.
