Is AI visibility just SEO with a new name?

Where is the sceptic right?

A large share of the work is not new, and pretending otherwise is how this category earns its bad reputation.

The inputs that make an engine able to name you are, in the main, things a competent search practitioner has been doing for years: accurate and consistent information about your business, being mentioned on sources other people read, pages that answer real customer questions in plain language, and a site the crawlers can actually reach. Rename none of that and it still works.

So when somebody tells you this is an entirely new discipline requiring an entirely new budget line, they are describing a sales structure rather than a body of work. The definitional comparison, including what genuinely differs in emphasis, is set out on GEO vs SEO and AEO vs SEO vs GEO, and this page will not restate it.

Where is the sceptic wrong, in one number?

At the point where a well-built site is assumed to produce a recommendation. It does not.

On 30 July 2026 we scanned 300 search-visible Australian businesses across six industries, put buyer-style questions to ChatGPT and Gemini, and ran automated checks on all 300 websites. Two figures from that scan sit directly on top of the rebrand argument.

Read the first two together. The thing classic technical SEO optimises hardest is in decent shape across the sample, and it barely predicts whether an engine recommends you. A correlation of 0.116 is not a weak relationship you can lean on. It is close to no relationship at all.

The sampling makes it sharper. Every one of the 300 was selected because it was already visible in classic search. These are businesses that have, by construction, won the old game. If AI visibility were SEO with a new name, they should have been winning this one too. Instead 41% of the 137 complete records were never named once.

That is the falsifying observation, and it is our own data producing it rather than a vendor's claim. It does not prove the two are unrelated. It does prove they are not the same thing.

What is actually different about the mechanism?

There is no list, and there is no position to hold.

A search engine returns a ranked set of documents that is broadly stable between two people asking the same thing. A generative engine composes an answer, and the composition varies. Across the 137 businesses with complete two-engine records, Gemini scored the same business higher than ChatGPT 77% of the time, ChatGPT was higher 5% of the time, and the average gap between the two engines for the same business on the same day was 25.3 points out of 100.

Two systems, one business, one day, a quarter of the scale apart. Nothing in classic search behaves that way, and it is why answers change between runs and why ranking language does not transfer. The unit of success is not a position. It is whether you were named, how often, for which questions, on which engine.

The second real difference is what feeds the answer. A page can rank on link equity and intent match without ever stating plainly what the business does, where it operates and who it is for. An engine composing a recommendation needs those facts to be legible and consistent wherever it reads them. That is why the work shifts towards writing pages that can be quoted and towards being described accurately by third parties, rather than towards ranking mechanics.

So is somebody charging me twice for the same work?

Sometimes, yes, and the r = 0.116 figure is how you check.

If a proposal is mostly schema markup, meta tags, an llms.txt file and page-speed work, then it is largely SEO deliverables with a new cover sheet, and our own dataset says that bundle barely predicts the outcome you are buying. Ask what in the scope addresses the things readiness does not cover: what third-party sources say about you, whether your business facts are consistent everywhere an engine reads them, and whether your pages answer the questions buyers ask out loud rather than the keywords they type.

One item is worth naming, because it gets sold hard. 71% of the 300 websites had no llms.txt file, which is a real and easily measured gap, and we report it as a legibility signal only, because Google has said it does not use llms.txt. A supplier presenting that file as the fix is not lying about the measurement but is overstating what it buys. The same test applies to schema markup.

Whether the remaining work is worth paying for at all is a separate question, argued from the same data on is AI visibility worth it.

How do I test the claim on my own business?

Put the two systems side by side on the same five questions and see whether they agree about you.

1. Write five questions a real buyer would say out loud. None of them may contain your business name, because a question that names you cannot test whether you get named unprompted.
2. Ask all five of ChatGPT and of Gemini. Record every business named in every answer, verbatim.
3. Search the same five in Google and record the top handful of organic results.
4. Compare the two lists.

If the names overlap heavily, then in your category the sceptic is close to right and your existing search work is carrying you. If you rank well and are absent from the generated answers, the two are demonstrably not the same thing for you, and no amount of terminology settles it either way. There is a longer version of this method on how to check what ChatGPT says about your business.

Our interest in the answer

Answerable sells AI visibility work, so we are the last people whose framing you should accept on trust.

That is why the number this page leads with is the one that cuts hardest against our own category. The full method, the per-industry splits and the raw rows are published so the arithmetic can be redone without asking us, and the scoring formula behind every figure is on The Receipts Engine. Had the correlation come out at 0.6, the honest version of this page would have said the sceptic wins. It came out at 0.116.

Settle it with your own data.

The free scan runs the comparison above for you: the questions, both engines, the answer text, and who was named instead of you. Then decide what to call it.