Goppa Research
AI visibility tools: what to check before you pay for one
An AI visibility tool answers one question: when someone asks ChatGPT, Google's AI Overview or Perplexity for what you sell, is your business named? Dozens of products now promise that answer, and they measure very different things under the same name.
We make one of these tools, so read this with that in mind. We wrote it to be useful even if you choose another one: it covers what a tool has to measure, what it cannot tell you, and what to ask before you pay.
A readiness score is not visibility
Many tools start, and some stop, at a score for your site: an llms.txt file, structured data, crawler access. Those are worth having. They are a prerequisite, not the result.
We measured 64 shops twice: once for technical readiness, and once by asking assistants real buying questions in each shop's category. 32 of the 64 scored a perfect 100 on readiness, and 25 of those 32 appeared in none of the answers. Across all 64, only 16 were named at all.
So a tool that only grades your site is grading the entry ticket. The number you are paying to learn is whether you get named, and that has to be measured by asking.
What a good tool measures
1. One assistant at a time
ChatGPT, Google's AI Overview and Perplexity search different indexes and cite different pages. A single blended score hides the one place where you are absent. Look for a number per assistant.
2. Counts, with the sample size
Assistants do not give the same answer twice. A useful figure says "named in k of n answers", from repeated runs of the same questions, rather than a bare percentage. Below a few dozen answers, a percentage implies a precision the sample does not have.
3. Questions you can read and change
The result depends entirely on what was asked. The questions should be the ones a buyer in your market would type, in their language, built from your own categories, and you should be able to see every one and replace the ones that do not fit.
4. Who is named instead of you
A zero is more useful with names next to it. Knowing which businesses the assistant recommends for your questions tells you what it considers a credible answer in your category.
5. The pages the assistant read
This is the part you can act on. To answer, the assistant opened a short list of pages: directories, reviews, comparison guides, forums. Getting onto those pages is most of the work, and some of them, such as directories with no gatekeeper, take an afternoon.
6. A change you can trust
When the number moves, the tool should compare the same questions on both dates and tell you whether the difference is larger than the day-to-day noise. Otherwise every small swing reads as a win or a loss.
7. A place, for Google
Google's AI Overview always answers from a location. A tool that reports a "global" figure for it has chosen a country without telling you. Ask which one.
Questions to ask any vendor
- How do you ask the assistant? Through an API with live search, or the consumer app? Both are legitimate, and they can answer differently; the vendor should say which, next to the number.
- How many times is each question run, and can I see every answer it is based on?
- What exactly counts as "named"? The brand in the text, a link to the site, or both?
- Do you separate real use from crawling? If the tool reads your server, a crawler indexing your pages is not the same as an assistant opening a page because someone asked. They should never be added together.
- What does one tracked question cost per month? It is the only unit that compares across vendors, because every plan bundles different things.
- Is there a free check I can run on my own site before I commit?
And what no tool can do
No tool controls which businesses an assistant names. A vendor that guarantees a position is selling a promise it cannot keep. What a tool can honestly offer is the measurement, the list of pages to work on, and a second measurement that shows whether the work moved anything.
Where Goppa fits
Goppa measures ChatGPT (through the OpenAI API with live search), Google's AI Overview and Perplexity, one at a time, as counts with the sample size. It shows the questions, who is named instead, and the pages each assistant read. Our WordPress plugin also counts real AI use on your site, kept apart from crawling. Since we also do SEO, the same account tracks your Google rankings and audits your pages. There is a free check, with no card.
The limits of what we just told you
The 64 shops are the set our own collection reached, not a census, and a different collection would produce a different set. 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. The checklist above is our view of what a measurement needs; other vendors may weigh it differently.
How to cite: Goppa Research (2026). AI visibility tools: what to check before you pay for one. https://trygoppa.com/answers/ai-visibility-tools — 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.