Goppa Research

The four layers between your business and an AI recommendation

"We guarantee you will appear in AI answers" cannot be said honestly, and the reason is structural rather than a matter of effort. There are four layers between a business and a recommendation. Two of them are entirely yours, one is partly yours, and the last one belongs to nobody.

Knowing which is which is what separates a promise you can keep from one that turns into a refund.

Layer 1 — Retrieval: being in the set at all

When an assistant searches, it searches an index. A page that is not in that index is outside the pool the answer is drawn from, however well it ranks somewhere else.

This is a gate, not an optimisation. Nothing above it matters until it is open. It is also the layer most tools skip, because it is not a file you can generate and hand over — it is a state you have to verify.

Yours, completely. Crawler access, indexing, and pushing new and changed pages instead of waiting to be crawled.

Layer 2 — Readability: what the page says to a machine

Structured data, identifiers, correct price and availability, a description with something an assistant can quote. This is the layer every product in this category sells first, ours included, because it is cheap to do and easy to verify.

Yours, completely. And not the constraint. We measured 64 shops on both readiness and appearance. The median readiness score of the shops that are never named is 100. For the shops that are named, it is 97. 32 of the 64 score a perfect 100, and 25 of those 32 appear in zero measured answers.

Do this layer anyway. It is a prerequisite and it costs little. Just stop expecting it to be the thing that moves the answer, because we have counted and it is not.

Layer 3 — Corroboration: what other pages say about you

This is where the number actually moves, and it is the half most people ignore because it is not on their own website.

We measured 81 agencies that sell AI visibility, using the question their own prospective clients would type. Not one of the 81 had its own website read as a source in its own measurement. The assistants answered by reading search results and other people's pages.

A technically flawless business that nobody writes about gives the model very little to lean on. That is not a marketing opinion; it is what the source lists look like when you count them.

Partly yours. The routes in are not equally open:

• Directories — you submit, you are in. No gatekeeper, and almost nobody bothers. Do these this week.
• Roundups someone else wrote — you ask to be included. Cheap, slow, and it works.
• Editorial coverage — a journalist decides. You can pitch; you cannot schedule it.
• Community threads — real participation only. Dropping links gets you removed, and it should.
• Encyclopedia entries — need independent coverage to exist first, so they are the last step, not the first. Anyone selling you one is selling you a deletion.

Layer 4 — Selection: the model's choice

Nobody controls this. Not us, not any agency, not the business itself. It is the same category of thing as "guaranteed first place on Google", and it should be treated with the same suspicion.

What can be said is that selection operates on what the first three layers put in front of it. You cannot make the choice. You can change what the choice is made from.

What can be promised, then

Everything on layers 1 and 2, because they are verifiable by you without trusting anybody's word:

• AI crawlers and the search crawler allowed — check your own robots.txt.
• Product pages in the index — search your own domain and count.
• New and changed pages pushed rather than waited on.
• Structured data complete, with identifiers — validate it in Google's own testing tool.
• Price and availability correct in the markup, so a €450 product does not appear at €38 because of an instalment figure.

None of those depend on a model choosing anything. What cannot be promised is the appearance itself — and a supplier who promises it either does not know about layer 4 or is counting on you not to.

Why this ordering is the useful part

Most effort in this category goes into layer 2, because it is the one with a checklist. The counting says layer 2 is not where the constraint is, and the source lists say layer 3 is. The work is harder there and the tooling is worse, which is exactly why the position is still open.

The limits of what we just told you

The 64 shops and the 81 agencies are the sets our own collection reached, not censuses, and a different collection would produce different sets. Assistants vary by day and by phrasing, and a business we record as never named may well be named on a question we did not ask. Layer 4 is, by construction, the part we cannot measure from outside: we can count what went in and what came out, and not why.

How to cite: Goppa Research (2026). The four layers between your business and an AI recommendation. https://trygoppa.com/answers/four-layers-between-you-and-an-ai-recommendation — CC BY 4.0.

No business is named for a poor result. We publish the counts and the method; the identity of individual businesses stays out of it.