Goppa Research
The readiness score does not predict anything. Here is what does.
We measured 64 online shops twice. Once for technical readiness, the way every tool in this category does it. Once for whether an AI assistant actually names them when someone asks what to buy.
The two have almost nothing to do with each other, and we can show the counts instead of the opinion.
The number that should end the argument
The median readiness score of the shops that are never named is 100. For the shops that are named, it is 97.
The invisible ones score higher.
32 shops score a perfect 100. 25 of those 32 appear in zero measured answers. 51 of 64 score above 90, and 16 of 64 are named at all.
If the technical layer were the constraint, this table would look different.
Why we went looking
Every product in this category, ours included, sells the same first step: publish structured data, serve an llms.txt, let the AI crawlers through. It is good advice. It is also cheap to verify, which is why everyone sells it.
What nobody was checking is whether it works. So we checked, on shops that are not our customers, and published the result including the part that is inconvenient for us.
What the assistants actually read
An assistant answering a buying question opens pages. We counted what kind, across the categories we measured:
- Roundups someone else wrote (31) — ask the author to include you
- Encyclopedia entries (28) — needs independent coverage first, so it comes last
- Directories you can join (23) — add yourself today, no gatekeeper
- Editorial reviews (17) — a journalist decides, so pitch or earn it
- Community threads (2) — real participation only; dropping links gets you removed
Read that list in order of difficulty rather than in order of size. The pages you can join without anyone's permission sit near the top, and they are the ones almost nobody bothers with. The hardest one — an encyclopedia entry — needs independent coverage first, so it is the last step, not the first.
The average is useless to you
In natural beauty, 1 of 12 shops is named. In supplements, 3 of 7. A shop that reads one industry-wide percentage and compares itself to it has learned nothing, because the answer depends on what it sells.
This is the argument for measuring your own category rather than buying a benchmark.
What we would do, in order
- Get the pages indexed. A page outside the index behind the assistant cannot be recommended, whatever its markup says. This is the step most tools skip, because it is not a file you can generate.
- Do the technical work anyway. It is necessary and it is cheap. It is just not sufficient, and the numbers above are how we know.
- Get on the pages the assistant already opens. Start with the ones that take no permission.
- Measure, then repeat. Not the score — the answer.
The method, and its limits
Buying questions were put to AI assistants with search enabled. Each answer was checked for the shop's domain. The readiness score comes from public checks run from outside, with nothing installed on the site.
Limits we will state before anyone else does: the sample is 64 shops, not a census; assistants vary by day and by phrasing; and a shop we record as never named may well be named on a question we did not ask. The full method and the sample sizes are at https://trygoppa.com/study
No shop is named here. The method promises aggregate reporting, and a study that embarrasses the shops it measured would not deserve to be believed twice.
Run the same check on your own domain, free and without installing anything: https://trygoppa.com
How to cite: Goppa Research (2026). The readiness score does not predict anything. Here is what does.. https://trygoppa.com/answers/readiness-score-does-not-predict-ai-recommendations — CC BY 4.0.
No shop is named. We publish the counts, the categories and the method; the identity of individual shops stays out of it.