We Verified 36 AI Visibility Tools. Google's AI Recommends Five of Them.

We spent this week reading 36 vendors’ pricing pages and recording every price, engine count and limit with the date it was read. Then we asked Google’s AI Overview which tool to buy, in the United States, Spain and France. It cited five of the 36.

Not five per market. Five in total, across all three.

The other 30 domains it cited belong to media, agencies, forums and companies that do not sell an AI visibility tool at all. And two domains appeared in all three markets: youtube.com and seranking.com.

How this was measured: Google AI Overview, one prompt per market, country and language set explicitly, cache bypassed, runs serial, all on 7 August 2026. Three runs in the United States, three in Spain, two in France. The French leg asked for four and two answered, which is the sample we have rather than the one we wanted. Cited domains only, never brand mentions. All three are in our measurement register with prompts and run counts.

Disclosure: EchoWi sells AI visibility measurement and is one of the 36, so this is a compilation by a company with an interest in the result. Everything needed to repeat it against us is here: the question, the surface, the markets, the run count and the dates, all stated below, and the rows are in our measurement register.

The short version

1. 35 distinct domains cited across three markets for the same question.
2. Two appeared in all three. YouTube and SE Ranking.
3. 30 of the 35 appeared in exactly one market.
4. Five of our 36 verified vendors were cited anywhere.
5. None of the market’s best-known names was cited at all.

What each market returned

The three questions are the same question in three languages: what is the best tool to track my brand’s visibility in ChatGPT.

Every domain, so you can check us

Publishing the counts without the membership would make this unrepeatable, so here is the whole set.

United States, 3 runs. Cited in all three: youtube.com, reddit.com, wpengine.com, seranking.com, dageno.ai, alhena.ai, promptrush.ai, mybrandi.ai. Cited once: brandlight.ai, eseospace.com, loopexdigital.com, jotform.com, columncontent.com, arcintermedia.com, leapshq.com, riffanalytics.ai, omr.com.

Spain, 3 runs. Cited in all three: youtube.com, acumbamail.com, blog.hubspot.es. Cited twice: surfeo.ai. Cited once: seranking.com, latevaweb.com, useomnia.com, amadora.ai, mybrandi.ai, ai-semantica.com, javadex.es.

France, 2 runs. Cited in both: seranking.com, invox.fr, youtube.com. Cited once: tool-advisor.fr, reddit.com, redback-optimisation.fr, fr.linkedin.com, embedsocial.com, abondance.com, dageno.ai, business.adobe.com, geoptie.com, dixmilleheures.fr, alambic.org.

The two domains that crossed every border

youtube.com was cited in every run of all three markets. That is consistent with everything else we have measured: in our count of twelve measurements across nine categories, YouTube appeared in eleven and was stable in eight. A software buying question in three languages does not change it.

seranking.com was cited in every run in the United States and France, and in one of three in Spain. It is the only vendor domain to appear in all three markets, and it is why we reviewed SE Ranking in the first place. Whatever else is true about that product, its pages are the ones this question retrieves.

Three more domains reached two markets: dageno.ai, reddit.com and mybrandi.ai.

Everything else, 30 of the 35, appeared in exactly one market. That is the same shape we found asking four surfaces one question, where 41 of 51 domains appeared on a single surface. Change the surface or change the country and you change the evidence.

Five of thirty-six

Here is the part that took a week to be able to say.

We read 36 pricing pages for this study, each price taken from the vendor’s own page on a stated date and recorded in our verified catalogue, which has grown since. Of those 36, the AI cited five:

Thirty-one were cited nowhere. That list includes every name this market talks about: Semrush, Ahrefs Brand Radar, Profound, Peec AI, Scrunch, Athena HQ, Writesonic, SISTRIX, Conductor, seoClarity, Evertune, Quattr and HubSpot’s new AEO product.

We are not claiming those products are bad. Several are the best-funded and best-known in the category, and two of them have the widest engine coverage we have verified. We are claiming something narrower and stranger: being the biggest vendor in a category does not get you into the answer to that category’s buying question.

Price has nothing to do with it

The five cited vendors run from €39 to $199 at their entry tier. The 31 uncited ones run from free to $2,500.

Sort the catalogue by price and the cited five appear at positions that look random, because they are: Surfeo at €39 is near the bottom, Alhena at $199 is near the top, and the three in between are unremarkable. Nothing about what a tool costs predicts whether the answer names it.

That is worth stating because it is the second half of a finding we published yesterday. Reading all 36 pricing pages at once showed that price barely buys engine coverage either: the cheapest tool in the catalogue tracks ten engines and a $499 one tracks five. So in this category the price tag predicts neither how much you get nor whether anyone will hear about you.

What the AI cites instead

Strip out the five vendors and the two platforms and what is left is the interesting part: media and agencies.

In Spain the stable set was youtube.com, acumbamail.com and blog.hubspot.es. An email marketing company and HubSpot’s Spanish blog, cited in every run, for a question about AI visibility tools.

In France the stable set was seranking.com, invox.fr and youtube.com. Invox is a content marketing agency. Below it, abondance.com and tool-advisor.fr, both French SEO media, and redback-optimisation.fr, an agency.

In the United States the stable eight included wpengine.com, a WordPress host, and reddit.com.

The pattern is the one we keep finding. A page that compares gets retrieved, whoever wrote it. We found it in flight booking, where every stable domain was a metasearch engine, in universities, where the stable sources sell accommodation and tutoring, and in project management, where zapier.com is cited in every run of a category it does not compete in.

Vendors publish product pages. Media and agencies publish comparisons. Only one of those answers “which is the best”.

What a vendor in that list of thirty-one should do with this

Stop optimising the product page for this question. It is not what gets retrieved. In three markets and eight runs, the domains cited in every run were a video platform, a forum, a hosting company, two email and content marketing blogs and one SEO suite.

Get into the comparisons that already rank. In France, three of the fourteen domains are French SEO media running “best AI visibility tools” pieces. Those are the pages the answer is built from, and they are reachable through ordinary press work.

Measure your own market, not the biggest one. Surfeo is cited in Spain and nowhere else. Omnia the same. A Spanish vendor measuring itself from a US tool would conclude it is invisible.

And check the answer exists before buying anything to watch it. This question produces an AI Overview in all three markets, which is why this study was possible. Two of five categories we measured in Spain returned no AI Overview at all across thirteen runs.

What this does not show

• Eight runs is not a survey. Three, three and two, on one surface, on one day. The French leg is two runs and every French number here should be read with that in front of it.
• One surface. Google’s AI Overview. ChatGPT, Gemini, Perplexity and AI Mode retrieve differently, and we have shown they cite almost nothing in common.
• One phrasing per market. A different wording of the same intent can return a different set; we have measured that too.
• Citation is not endorsement, and absence is not a verdict. A tool cited here is one whose pages answer this question well. A tool absent may be excellent and simply not written about.
• Nothing causal. We changed nothing and measured what came back.
• We are in the data. EchoWi sits in the same catalogue and was measured on the same terms as every other tool here. Nothing was exempted, and we sell in this category, which is why the prompt, market and run count are published for every figure.

Common Questions About What AI Recommends

Which AI visibility tools does Google’s AI recommend?

In our measurement on 7 August 2026, across three markets and eight runs, five vendors from our 36-tool verified catalogue were cited: SE Ranking in all three markets, Dageno in two, and Alhena, Surfeo and Omnia in one each. Most of the domains cited were not vendors at all.

Why are the biggest AI visibility tools not cited?

We can report the pattern, not the cause. The domains cited in every run tend to be pages that compare products rather than pages that sell one, and comparisons are mostly written by media, agencies and adjacent companies rather than by vendors. Thirty-one of the 36 tools we verified, including the best-funded names in the category, were cited in none of the eight runs.

Does the answer change between countries?

Almost entirely. Thirty of the 35 domains cited appeared in exactly one of the three markets. Only youtube.com and seranking.com appeared in all three. If you sell in more than one country, a measurement from one of them tells you very little about the others.

How many runs do I need to measure this myself?

More than one, always. In this study eight runs produced 35 domains of which 30 appeared once, so a single run would have given you a confident and largely accidental list. How much repetition a category needs runs from about three to about thirty depending on how volatile it is.

Is EchoWi in this catalogue?

Yes. We sell in this category, and our own pricing page was read on the same day and to the same standard as the other thirty-five. Nothing was exempted from the sweep and nothing was exempted from the measurement. The prompt, the market, the run count and the date are published for every figure here precisely so the comparison can be repeated against us.

Can I reproduce this?

Yes. One prompt per market, country and language set explicitly, cache bypassed, three runs where the provider allows it, on Google AI Overview. The prompts, markets, dates and run counts are in our measurement register, and the 36 prices they are compared against are in our verified catalogue with the date each was read.