GEO is the work of getting named inside an AI answer. SEO is the work of ranking a link on the page underneath it. On the same query those are now two different result sets. Across twelve UK search results measured on 10 August 2026, 46 of the 85 sources Google’s AI Overview cited did not appear in the organic top ten below it. That is 54%. Counted the other way, 58 of 99 organic results had no citation in the AI Overview sitting above them.
So the useful answer to “is GEO just SEO” is that about half of it is. The half that is not is where the disagreement in this subject comes from.
This count is the evidence underneath answer visibility, which sets out what the whole subject consists of and what else has been measured against it.
How was this measured?
By reading both halves of the same page.
Method. Ahrefs serp-overview, GB market, pulled on 10 August 2026, for twelve queries in the AI search vocabulary. For each query we took the URLs Google displays with its AI Overview (Ahrefs types these rows ai_overview_sitelink) and the organic results (organic), then matched the two lists. The headline figure matches on exact URL. A looser match, counting an AI Overview source as ranking if any page on its domain sits in the top ten, is reported alongside it.
Three limits, stated because they bound everything below.
Ahrefs snapshots each SERP on its own schedule. The twelve here were crawled between 17 July and 7 August 2026. The crawl date is given per query in the table. One of them, “generative engine optimisation services”, was last crawled on 11 May and is three months stale. It is kept in and flagged rather than dropped. The figures are restated without it.
The links Google shows with an AI Overview are not necessarily every source it used. Google does not publish the full retrieval set, so this measures displayed citations, which is the only part anyone outside Google can count.
Organic counts run from seven to nine rather than a flat ten, because SERP features occupy positions and Ahrefs returned fewer organic rows on some pages. The percentages are computed on the rows returned.
One query was excluded outright. On “generative engine optimisation”, Ahrefs returned Google redirect URLs for the AI Overview sources and root or truncated URLs for several organic results, so the two lists cannot be matched against each other. It is named here rather than quietly left out.
Which queries were tested and what did they show?
| Query | SERP crawled | AI Overview sources | Organic results | Sources also ranking |
|---|---|---|---|---|
| chatgpt seo | 17 Jul 2026 | 15 | 8 | 6 |
| is seo dead | 18 Jul 2026 | 10 | 7 | 3 |
| ai visibility tools | 18 Jul 2026 | 7 | 9 | 3 |
| generative engine optimisation agency | 28 Jul 2026 | 8 | 9 | 3 |
| ai overviews | 28 Jul 2026 | 8 | 8 | 2 |
| geo agency | 31 Jul 2026 | 4 | 8 | 4 |
| ai visibility | 1 Aug 2026 | 6 | 9 | 2 |
| llm seo | 2 Aug 2026 | 8 | 8 | 5 |
| answer engine optimisation | 2 Aug 2026 | 5 | 8 | 4 |
| aeo vs seo | 4 Aug 2026 | 3 | 8 | 3 |
| geo vs seo | 7 Aug 2026 | 8 | 8 | 4 |
| generative engine optimisation services | 11 May 2026 | 3 | 9 | 0 |
| Total | 85 | 99 | 39 |
Thirty-nine of 85 cited sources rank in the organic top ten on the same query, so 46 do not. Allowing a domain match rather than an exact URL lifts it to 41, which still leaves 44 citations with no page-one presence at all.
Dropping the stale May snapshot moves the headline from 54% to 52%. The finding does not depend on it.
Every one of the twelve SERPs carried an AI Overview. On eleven it occupied position one. On “ai visibility tools” a single organic result sat above it.
How much of an AI Overview comes from page one?
Under half. It varies more than the average suggests.
The tightest coupling is “geo agency”, where all four cited sources also rank. The loosest, setting the stale snapshot aside, is “ai visibility”: six sources cited, two ranking. The AI Overview there leans on a Semrush index subdomain, a glossary page on llmpulse.ai and an Onclusive blog post, none of which hold a top-ten position on the query.
This is the part worth carrying into a planning conversation. A page-one ranking is neither necessary nor sufficient for being named in the answer above it. On these twelve queries it coincides with citation slightly less than half the time.
Why is YouTube cited four times more often than it ranks?
Because the two systems weight video very differently. This is the clearest single difference in the dataset.
Thirteen of the 85 AI Overview citations, 15%, point at youtube.com. YouTube appears in the organic top ten four times across 99 results, 4%. Same twelve queries, same day, a near four-fold gap.
On “chatgpt seo” the AI Overview cites four separate YouTube URLs, including two shorts, while no YouTube result ranks in the organic ten at all. On “ai visibility tools” it cites two videos and no video ranks.
We are not going to claim we know why. What can be said is narrower and still useful: on this sample the AI Overview reaches for video at roughly four times the rate the organic results do, so a market where nobody has made the video is a market where the citation is cheap.
What else shows up on both lists?
The domains cited most often are not the ones you would predict from the rankings.
| Domain | AI Overview citations | Organic appearances |
|---|---|---|
| youtube.com | 13 | 4 |
| semrush.com | 5 | 5 |
| neilpatel.com | 4 | 3 |
| reddit.com | 3 | 6 |
| impressiondigital.com | 3 | 2 |
Reddit is the interesting row. It ranks twice as often as it is cited, which is the reverse of YouTube. Forum threads win the link and lose the answer. Video wins the answer and loses the link.
Semrush is the only domain in the set that draws level, cited five times and ranking five times across twelve queries. It is also the largest content estate in the sample, which is consistent with what we found comparing the AI visibility tools: citations accrue to big, long-established domains. The same mechanism appears to run underneath both halves of the page.
What actually differs between the two jobs?
Four things, on the evidence above rather than on principle.
- The unit you are competing for. SEO competes for one of ten link slots. GEO competes for a sentence in a paragraph that may name three brands or none. There is no tenth place in an answer.
- The formats that qualify. A video that would never rank on a commercial query gets cited on one. So does a glossary entry on a domain with no rankings.
- Where the work sits. A ranking is won mostly on your own site. A citation is frequently won on someone else’s, because the pages a model reads are often lists, threads and comparisons written by third parties.
- What measurement looks like. Positions are ordinal, dated and reproducible. Citations are probabilistic, undated by the platform and different on the second run. Five methods and what each one hides covers this properly.
What does not differ
More than the category’s marketing suggests.
Roughly half the AI Overview is drawn from pages that already rank, so the technical and editorial work that earns a ranking still earns citations about half the time. Nothing in this dataset supports abandoning it. Nothing supports treating it as sufficient either.
Crawl access sits underneath both. It is also the cheapest thing to get wrong. Ten UK agencies, all allowing every OpenAI crawler, hold ChatGPT citation counts running from 0 to 92, which is the same shape of finding: the entry ticket is not the race.
What this does not tell you
It is twelve queries in one market on one platform, all in the AI search vocabulary. A market with different SERP furniture will produce different overlap. Nothing here should be read as a general figure for UK search.
It measures displayed citations, not retrieval. If Google reads a page and does not link it, this method records nothing.
And it is a single snapshot per query. Citation sets have been observed to change within a day, so the honest reading of 54% is that it is the number on these twelve pages on the dates given, not a constant.
For the vocabulary underneath the comparison, what generative engine optimisation actually is and what answer engine optimisation is define the terms. For the same measurement applied to your own market, the afternoon method costs nothing but time.
Data in this article: Ahrefs serp-overview, country=gb, pulled 10 August 2026 for twelve queries. Ahrefs SERP crawl dates run 17 July to 7 August 2026 and are given per query in the table. Matching performed on exact URL, with a domain-level match reported alongside. The “generative engine optimisation” SERP was excluded because Ahrefs returned redirect and truncated URLs that cannot be matched; the “generative engine optimisation services” SERP is three months stale and the headline is restated without it.