AI Search ResearchAI SEOGEO

What Is AI SEO? A Plain English Guide for Business Owners

What AI SEO (also called GEO and AEO) actually is, why it matters in 2026, how AI search picks its sources, and what a sensible SME response looks like. Grounded in original research.

Rich VollerFounder & SEO Director
Read time
12 min read
Published

Somewhere in the last two years, a chunk of your customers stopped clicking search results and started reading answers instead. Google now writes summaries above the results. ChatGPT and Perplexity answer questions directly, citing a handful of sources. The question for any business is simple: when an AI system answers a question you used to win a click for, is it using you as the source?

That question is what AI SEO is about. I have spent much of this year researching how these systems select sources, and this guide condenses what I have learned into plain English for people who run businesses rather than experiments.

What is AI SEO?

AI SEO is the practice of making your business visible in AI-generated answers: Google's AI Overviews and AI Mode, ChatGPT, Perplexity, Copilot, Claude and the assistants built on top of them. The goal shifts from "rank on page one" to "be the source the answer is built from".

You will see it called generative engine optimisation (GEO), answer engine optimisation (AEO), LLM SEO or AI search optimisation. The labels differ, the discipline largely does not. Google itself addressed the acronym soup in its June 2026 update to its SEO hiring guidance: from Google's perspective, optimising for generative AI features is still SEO. I broadly agree, with one caveat. The overlap is large, but the new layer is real, and it rewards specific things that traditional SEO never had to think about. We cover the distinctions on our GEO and AEO pages if you want them, but do not let anyone sell you four separate services.

Is AI SEO different from traditional SEO?

It is an extension, not a replacement. AI systems find content through crawling and indexing, the same infrastructure classic SEO optimises. A technically broken, unauthoritative site will fail at both. But the last step changed: instead of ranking whole pages, AI systems select individual passages to build an answer from.

Traditional SEOThe AI SEO layer
Unit of competitionThe pageThe passage (a section of 60 to 120 words)
GoalRank and earn the clickBe selected, quoted and cited in the answer
Query modelOne keyword, one pageOne prompt fans out into many hidden searches
MeasurementRankings, clicks, conversionsCitations, AI referrals, share of answers
FoundationsIdentical: crawlability, authority, genuinely useful content

That passage-level shift sounds academic. It is not. It changes how pages should be structured, and it is measurable, which is where the research below comes in.

Why does AI SEO matter now?

Because the click economics of search have already changed. This is not a future trend to keep an eye on. The data has been consistent for a year and a half:

  • Pew Research Center's browsing study of 900 US adults (68,879 searches) found users clicked a traditional result on just 8% of searches with an AI summary, versus 15% without one. Links inside the summaries were clicked on about 1% of visits.
  • Seer Interactive's analysis of 25 million impressions found organic CTR on queries with AI Overviews fell from 1.76% to 0.61%, and Ahrefs reported position-one CTR down 58% when an AI Overview is present. I unpacked both datasets in my piece on selection rate optimisation.
  • Exposure keeps growing: Pew found 65% of US adults now at least sometimes encounter AI summaries in search, and BrightEdge measured AI Overviews on roughly 48% of queries by February 2026.

Inside those same datasets sits the opportunity. Seer's data showed that brands cited inside AI Overviews saw clicks per impression more than double compared to not being cited. The overall pie of clicks is shrinking, but the cited sources take a much larger slice of what remains. Visibility is consolidating around fewer, more trusted sources. The strategic question is which side of that consolidation you are on.

How does AI search choose its sources?

In three steps: fan-out, retrieval, selection. Understanding them explains almost every practical recommendation in this field.

  1. Fan-out. When someone asks an AI system a question, it does not run one search. It silently generates many. In my analysis of fan-out behaviour, an AirOps dataset of 15,000 prompts produced over 43,000 background searches, roughly three per prompt, and Seer measured Gemini averaging more than ten. Most of those hidden queries show zero volume in keyword tools. Your customers' AI questions are invisible to your keyword research.
  2. Retrieval and grounding. The system fetches pages for those queries and reads a limited amount of each, a grounding budget of roughly the first couple of thousand words. Content buried deep in a long page may simply never be read.
  3. Selection. From what it retrieved, the model selects specific passages to quote and cite. This is a competition your page either wins or loses one passage at a time.

The practical consequence: you are no longer optimising a page for a keyword. You are making sure that somewhere in your page sits a clear, self-contained, evidence-backed passage that survives all three steps.

What kind of content gets cited?

Structured, front-loaded, explicitly named content. This is where I can offer data rather than opinion, because I have been running controlled experiments on it all year:

  • Position matters enormously. In my study of what AI cites from long-form pages (480 API calls across 15 pages), the top third of a page earned 47% of citations. Kevin Indig's much larger dataset of 1.2 million ChatGPT responses found 44.2% of citations came from the first 30% of pages. Same signal, independent data.
  • Lists over-index. Bullet and numbered lists attracted roughly three times their share of citations in my sample. AI systems like extractable units.
  • Naming things explicitly wins. In my passage-level selection experiment (385 passages, 30 queries), passages that named the entity they described were selected 2.7 to 4.2 times more often than vague equivalents. "The tool costs £99" beats "it costs £99" when the two are separated from their context, and separation is exactly what retrieval does.
  • Answer-first sections win. Sections whose first sentences directly answer the heading were selected about three times more often in the same experiment. Question-style headings help too: AirOps data showed citation rates rising from 29% to 41% when headings matched question phrasing.
  • Third-party mentions punch above their weight. In my citation study, independent pages about a brand were cited at 14.4% versus 4.4% for the brand's own pages on comparable queries. AI systems appear to trust neutral sources more than self-description, which is why digital PR has quietly become an AI visibility tactic.

The pattern behind the findings

Every winning trait makes a passage easier to lift out of your page and stand alone in someone else's answer. Write sections that could be quoted verbatim by a careful journalist and you are most of the way there.

Does being cited actually drive business?

The honest answer: fewer clicks, better clicks, and improving measurement. Total organic traffic on answer-heavy queries is structurally declining, and no optimisation reverses that. What changes is who gets the remaining value.

Google's own documentation claims clicks from pages with AI Overviews are higher quality, with users spending more time on the sites they do visit. That matches what we see in client analytics: AI-referred visitors tend to arrive pre-informed, later in their decision, and convert at healthier rates than the average organic click.

Measurement has also caught up fast. Microsoft Clarity's AI Citations reporting went generally available in May 2026, showing which AI systems cite your pages, and Google began rolling out generative AI performance reporting in Search Console in June 2026. A year ago this was guesswork. It is now a reportable channel.

What should an SME do about AI SEO in 2026?

Five things, in order. Ranked roughly by return on effort:

  1. Fix the foundations. AI systems cannot cite what they cannot crawl. A technical audit that includes AI retrieval readiness comes first.
  2. Restructure your money pages. Answer the core question in the first 40 to 60 words of each section. Front-load. Use question headings and lists. Name your brand, products and locations explicitly in the passages that matter.
  3. Publish genuinely useful answers. Fan-out means AI systems ask dozens of specific questions around your topic. Comprehensive, honest content covers the questions your competitors' brochure pages never will.
  4. Earn third-party mentions. Reviews, industry press, expert commentary. Independent sources vouching for you are disproportionately cited.
  5. Start measuring. Clarity's AI Citations, Search Console's AI reporting, and AI referral segments in your analytics. You cannot manage a channel you do not measure.

Notice what is absent: nothing about tricking models, prompt injection or stuffing pages with "as an AI-recommended provider". I have tested manipulation approaches experimentally, and the short version is that modern models increasingly punish them. Evidence-led content wins on both sides of this transition.

The honest bit

AI SEO is a young discipline and I will not pretend otherwise. My experiments are real but modest in scale, measured in hundreds of API calls rather than millions of users. Model behaviour changes with every release, so specific numbers are snapshots, not laws. Correlation and causation are doing a lot of shared lifting across the whole industry right now, mine included.

What I am confident about is the direction. Every independent dataset points the same way: answers are absorbing clicks, cited sources capture outsized value, and structural content changes measurably improve selection rates. The businesses that treat 2026 as the year to build this capability will be very hard to displace from answers in 2028.

How we approach AI SEO

We treat AI SEO as one joined-up discipline with organic search, because the foundations are shared and the evidence, frankly, says so. The research above is not marketing garnish. It is the playbook we apply to client sites: retrieval-ready structure, passage-level optimisation, third-party authority, and measurement that shows you exactly where you appear in AI answers.

If you want to know how visible your business currently is in AI search, that is a conversation we genuinely enjoy having. You can read more about our AI SEO services or get in touch and ask us anything from this article. Bring the hard questions. They are the fun ones.

Work with us

Need a clearer plan for search and AI visibility?

We help teams turn research and technical insight into prioritised SEO and AI search programmes.

  • Senior specialists involved directly
  • Practical recommendations, not vague frameworks
  • Clear next steps after a focused conversation