Is it all “just SEO”? Or are there new strategies, channels, or tactics you should be prioritizing?
You may not know how to measure or report on everything, but that hasn’t changed the demands from the C-Suite to deliver performance in the same data-driven methods that they’ve become accustomed to over the years.
An approach to AI Search requires education and flexibility, since it’s fundamentally different from the way the SEO industry has approached strategy and performance reporting.
Users never wanted 10 blue links in the first place. They wanted Star Trek technology that can do what we ask and answer questions or solve problems immediately. AI can eventually take us there if done correctly, and we need to be ready.
iPullRank’s own Mike King talked about the history and future of search, how we as an industry can prepare, and some helpful tips from our AI Search Manual in our recent webinar, “How to Win Visibility in AI Search: The AI Search Manual Walkthrough.”
“SEOs need to evolve,” he said. “There’s a lot more that we need to understand as a community.”
Here are some highlights from the webinar in case you missed it.
The Evolution of Search
When AI Overviews launched in May 2024, they had a huge impact on traffic and clicks. Fewer people were clicking over to websites because they now had a summary of information to answer their questions. However, this also led to more qualified leads and more conversions.
“Traffic from large language models or AI search is becoming more valuable,” Mike said.
With this change, we need a new way of strategizing, and that way is generative engine optimization (GEO), a name Mike says we should be using from now on because the publishing industry has already dubbed it that.
iPullRank has its own strategy called Relevance Engineering (r19g), which is the intersection of AI, information retrieval, digital PR, user experience, and content strategy. As users relearn how to search in a multimodal environment of voice, video, and image search, r19g will help us adapt.
“Users are realizing the more specific the information is in their query, the more specific the response will be.”
Mike emphasized that we need to rely mainly on “statistical rigor” and try to obtain as many data points as we can get our hands on. Facts, statistics, and accurate information have always been valuable, but in a world of growing AI slop they’re even more important.
Google Defends its Title
Google is still the most dominant force in search, and Mike thinks they will eventually win the AI battle.
Despite the results of the Department of Justice antitrust trial against Google, the giant still has the proprietary data advantage with an active user base of over one billion. And even though the remedies demand that they hand over that data to competitors, after they go through the inevitable appeals, Google may even change what that data looks like. There are many ways they could get out of this unscathed.
They also have access to first-party chip manufacturing and don’t need to rely on NVIDIA, which gives them another advantage.
The Problem with AI
According to a study done by Ahrefs, 86.5% of top-ranking pages now include some form of AI assistance. And this AI usage does not impact rankings.
You read that right. Just because content is AI-generated does not mean it won’t rank well.
Hallucinations are still not solved, though, so this doesn’t bode well for the future of content unless they can figure out how to solve that.
A Columbia Journalism Review article found that for a controlled information retrieval task, chatbots collectively provided inaccurate or misleading answers more than 60% of the time, nearly always without acknowledging uncertainty.
As Mike said during the webinar, we need to reconsider what we’re doing and identify all of the gaps.
“Now is the time to shed all the things that no longer serve us that we inherited in the SEO industry.”
The Rise of AI Overviews and Semantic Search
Google’s launch of AI Overviews and AI Mode marks a shift from deterministic rankings, where content is surfaced largely as-is based on predefined scoring signals, to probabilistic rankings driven by large language models that interpret, synthesize, and reason across multiple sources.
At the heart of most AI search platforms is retrieval-augmented generation (RAG). By grounding in fresh, externally retrieved data, RAG addresses the fundamental weaknesses of large language models: hallucinations and knowledge cutoffs. RAG systems can deliver answers that are both fluent and factual.
Search has also moved from lexical to semantic. Lexical search is driven by the distribution of words, while semantic is driven by a representation of meaning.
Hybrid retrieval combines both lexical and semantic search then re-rank the documents using methods like reciprocal rank fusion (RRF). However, 95% of SEO tools still only perform lexical analysis.
Google’s next leap was to embed everything it cared about in the search ecosystem to create a unified semantic framework where any object, website, author, entity, or user profile could be compared to any other in the same high-dimensional space.
These embeddings allow Google to “remember” patterns and tailor results in real time, shaping which sources are selected and how answers are framed, making search more personalized.
AI Overviews gave us one of our first tastes of semantic search. They generate an answer and look for documents that can corroborate what was said in the answer.
AI Mode is similar, but its query fanout uses a longer pipeline of the same idea. There are more reasoning stages that go into it, and more passages and pages considered for more queries.
Unlike traditional search engines like Google and Bing, LLMs like ChatGPT and Perplexity don’t “crawl,” they request, and they don’t index content or cache the pages so they are requesting in real time.
The query process for LLMs is a five-step process:
- Expansion: This is the stage where the system broadens the scope of what it is looking for, aiming to cover both the explicit and implicit needs behind the request.
- Routing: This is where the fan-out map becomes operational. Each sub-query is now a small task in its own right, and the system must decide which source or sources can best satisfy it, which modality is most appropriate for the answer, and which retrieval strategy will be used to get it.
- Retrieval: Content must match the expected modality or it won’t be retrieved. Ensure multi-modal parity (text, structured data, transcripts, etc.), place content where the routing logic looks (e.g. API-friendly formats, transcripts for procedural content), and align content with routing profiles to increase retrieval across fan-out branches.
- Selection: This often applies harm and safety filters that draw on both explicit policies and learned patterns from training data.
- Synthesis: High-quality content can be excluded from synthesis if it isn’t easily extractable. Interactive designs that aren’t crawlable or long-form narratives that bury key facts risk being skipped in favor of denser, more accessible material.
In this new retrieval-and-synthesis pipeline, the query you type is not the query the system uses to gather information. Instead, the initial input is treated as a high-level prompt that sets off a much broader exploration of related questions and possible user needs. The system decomposes the query, rewrites it in multiple forms, generates speculative follow-ups, and routes each variant to different sources.
Getting Visibility in AI Search
The fight for visibility in AI-driven search is about making content easy to extract, interpret, and reuse. If your content is buried in long, unstructured paragraphs, chances are it won’t surface. Clear headings, concise paragraphs, and scannable lists dramatically increase the odds of being pulled into an AI-generated response.
But visibility goes beyond your website. AI systems pull from the full content ecosystem:
- Audio clips
- Video transcripts
- Social media
- User-generated content
- Digital PR mentions
If your strategy is confined to a single channel, you’re basically accepting “a one-in-four chance” of surfacing. Omnimedia strategies expand your footprint so that wherever the AI looks, it finds you.
Our approach to securing that visibility at iPullRank starts with a content audit, followed by semantic and latent intent research to ensure alignment with how people search and how AI interprets meaning. We then structure and augment content that’s entity-rich, embedding-friendly, and supported by qualifiers, locations, and related entities.
Other semantic practices can also help, including:
- Structured data
- Layout-aware HTML chunking
- Writing for synthesis
- Using semantic triples
- Strategic use of numbers and data in your content
It’s more of a challenge to measure KPIs in the AI age, so we need to rethink what we’re measuring and how. Measurement now happens across three buckets:
- Input metrics: How much high-quality content is produced and distributed
- Channel metrics: How well content performs on each platform
- Performance metrics: The ultimate business outcomes
Continuous testing and iteration ensure the strategy stays effective as search evolves.
And now that we have the embeddings that OpenAI and Gemini use, we can simulate rankings based on the changes made (like in our proprietary program Qforia).
“We can effectively simulate what a change is going to do from how they’re measuring relevance against these prompts and queries,” Mike said.
It’s Not Just SEO
AI search forced a big change in how information is retrieved, interpreted, and surfaced. We need to adapt quickly, structure content so it’s easily extractable, and show up across the full ecosystem where AI is pulling answers. That means embracing GEO, expanding your team in new ways, applying semantic practices, and treating visibility as omnimedia.
The industry is at a point where old tactics no longer serve us. We can’t rely on rankings alone, and we can’t expect AI systems to do the work of parsing through messy or inaccessible formats. Clean layouts, entity-rich language, structured data, and supporting assets across video, audio, and PR are what make content usable.
As Mike King pointed out, the real opportunity is to combine statistical rigor with creative execution. We have the tools now to simulate rankings against LLMs, test strategies before deploying them, and iterate in a way that matches the speed of search innovation. That’s how we’ll stay ahead.
But until everyone gets to that point, Mike believes we’ll remain stuck:
“Until we see our software catch up, what everyone is talking about is an idealistic form of SEO that most people aren’t doing.”
