The Tools Google's AI Ignored Get 151,387 Searches a Month. The Ones It Named Get 7,864.

We published a study today on which tools Google’s AI recommends when you ask it what to buy. Then we looked up how many people search for each of those names. The answer it gave and the answer the search box gets are almost unrelated.

Fifteen of the tools the AI never mentioned draw 151,387 searches a month in the United States between them. The three it did mention that have any measurable demand draw 7,864.

That is 19 times more demand on the side the answer ignored.

How this was measured: search volumes from our keyword research tool on 7 August 2026, United States, English, with clickstream disaggregation switched on so that grouped close variants are separated. The citation data is the companion study published the same day: Google AI Overview, one prompt per market, country and language set explicitly, cache bypassed, three runs in the United States, three in Spain, two in France, all on 7 August 2026. Cited domains only, never brand mentions.

Disclosure: EchoWi sells AI visibility measurement, so we are one of the tools in this catalogue and one of the ones the AI did not cite. We are not in the demand table below, because we do not publish our own numbers. Every figure here belongs to somebody else.

The short version

1. The two largest brands in the category by search demand were cited in none of the eight runs. Semrush draws 83,144 a month. Ahrefs draws 39,937.
2. Three of the five vendors the AI did name have essentially no brand demand. One has none we can measure at all.
3. The ratio is 19 to 1 against the tools that were cited, and it survives at 17.9 to 1 after we throw out the two names we cannot trust.
4. The exception proves it is not an inverse law. SE Ranking was cited in all three markets and does have real demand, 7,604 a month.
5. Being searched for and being cited are two different distribution channels. Winning one predicts almost nothing about the other.

The fifteen the answer never named

Thirty one tools in our verified catalogue were cited in none of the eight runs. These are the fifteen of them whose names a buyer in this category would recognise, with the monthly United States search volume for the query that buyer would actually type.

Eight of the fifteen clear a thousand searches a month. The total is 151,387.

One row in that table has already survived this treatment. Quattr’s raw figure is 27,100 a month, which would make it the most searched tool in the category by a distance. Disaggregated, it is the 627 shown above. We took that number apart yesterday, along with a second tool that fell from 9,900 to 53, and it is why clickstream separation is switched on for every volume in this piece.

Two of those numbers are not what they look like

Before anyone repeats the total, here is what is wrong with it.

evertune is a guitar bridge. EverTune makes a self-tuning bridge for electric guitars, and it has been on the market far longer than any AI visibility tool. The tell is not the name, it is the price of the click: $1.86. Every genuine business tool on this list that anyone bids on sits between $9 and $26. Peec AI is $21.84, Otterly is $22.78, Scrunch is $18.89. A dollar eighty-six is a consumer query. We are not going to claim those 3,600 searches for a software company.

profound ai is shared with a medical imaging product. ProFound AI is a breast-screening tool from a different company entirely. Here the tell does not save us: the click costs $12.31, which is squarely inside the range a B2B software query would command, so the two products are genuinely mixed in that number and we cannot separate them. Treat 6,600 as a ceiling, not a measurement.

Take both out and thirteen unambiguous names remain, drawing 141,187 searches a month. The finding does not need the two doubtful ones:

141,187 ÷ 7,864 = 17.9

The five the answer did name

Add the ones we can measure in the market where they were cited and the whole cited side comes to 7,864 in the United States, plus 90 in Spain.

SE Ranking is the exception, and it matters. It was the only vendor cited in all three markets, and it is also the only cited vendor with a real brand behind it: 7,604 a month in the United States, 4,379 in Spain. So this is not an inverse law where obscurity earns citations. It is worse than that for anyone planning a budget: the two variables barely relate at all. One tool has both. Three have citations and no demand. Fifteen have demand and no citations.

Three of the five do not have names, they have words

This is the part that makes the demand side hard to measure honestly, and it is worth showing rather than hiding.

alhena returns 720 searches a month with informational intent and no cost per click at all. Alhena is a star, the brightest in Gemini after Castor and Pollux. Those are astronomy searches. The company’s actual query is alhena ai, which returns 260 searches at $19.03 a click. The second number is a hundredth of the first in volume and infinitely more valuable per visit.

omnia returns 5,335 searches a month in Spain at €0.34 a click, informational. Omnia is Latin for “everything” and a common brand name across a dozen industries. The company’s query is useomnia: 90 searches at €24.27.

surfeo returns 70 searches in Spain with no cost per click. In Spanish, surfeo is the first person singular of the verb to surf.

The pattern is the same each time. The cost per click tells you which query the buyers are on, and the volume tells you how few of them there are. Thirty four cents against twenty four euros is not a rounding difference, it is two different populations using the same string. Any study that adds up the big numbers here is counting stargazers, Latin students and surfers.

Why this probably happens

We can report the pattern. We cannot prove the mechanism, and we are not going to pretend a correlation of this size in one measurement is a law.

The most plausible reading is that the question is not about the brand. Someone asking an assistant for the best tool to track their visibility in ChatGPT is asking a category question, and the pages that answer category questions are comparisons. Comparisons are mostly written by media, agencies and adjacent companies, not by the vendors themselves. That is what the citation data showed: strip out the five vendors and the two big platforms and what remained was an email marketing company, a WordPress host, a content marketing agency and two SEO trade publications.

Brand search volume measures something different. It measures how many people already know the name and go looking for it. A company can be the most searched name in its category and still have published nothing that answers “which is best”, because answering that question means writing about your competitors.

That is a hypothesis, and it is testable. It is not a finding, and we have not tested it.

What we would do with this if we were in the fifteen

Stop reading brand search volume as a proxy for AI visibility. It is a measurement of a different channel. On this evidence it is close to uninformative about the other one, and it is the number most marketing teams already have on a dashboard.

Do not conclude your category is safe because you are the biggest name in it. Semrush and Ahrefs are the two most searched names on this list by an order of magnitude and neither appeared in any of the eight runs.

Look at what got cited instead of you. In every market we have measured, the stable citations were pages that compare things. That is a content gap with a known shape, and it is reachable.

Measure the market you sell in. Surfeo and Omnia were cited in Spain and nowhere else. A Spanish vendor watching a United States dashboard would conclude it is invisible.

The strangest row replicated the next day

The table above rests on one day. On 8 August 2026 we ran the United States prompt again, same string, same surface, cache bypassed, three runs:

dageno.ai was cited in all three runs again. It is the most awkward row in the table above, a domain with nothing measurable behind its name in either market where it was cited, and a day later it is not merely present but present every time. The one-session explanation is gone.

workduo.ai joined it, in two runs of three. Its name draws about 30 searches a month in the United States, which is the same story one order of magnitude up from nothing.

And the counter-example held. seranking.com, the single tool in this study with both real demand and a citation, 7,604 searches a month, was also in all three runs. So this is not a rule that demand predicts absence. It is that demand predicts nothing in either direction, which is the harder claim and the one the second day supports.

One thing the second day does not support is stability. The same prompt at the same three runs gave eight stable domains of seventeen on 7 August and five of fourteen on 8 August. The membership of the cited set is not fixed, so the right reading of any single row here is one observation, and we have published what happens when we re-run our own stability study for the same reason.

What this does not show

• Eight runs is not a survey. Three in the United States, three in Spain, two in France, one surface, one day. Every conclusion here inherits that.
• One surface. Google’s AI Overview. We have shown that four surfaces asked the same question cite almost nothing in common.
• Some of these are parent brands, not products. semrush and ahrefs measure demand for entire SEO suites, and the AI visibility module is one feature inside each. The same is true to a lesser degree for Surfer, Writesonic, SISTRIX, Conductor and seoClarity. Pure plays like Peec AI and Scrunch have nothing else in the name. So the top of that table overstates demand for the specific product. We first wrote that this widens the gap, and that was reasoning about one side of a ratio. SE Ranking, the vendor carrying 7,604 of the cited side’s 7,864, describes itself on its own homepage as AI SEO software, which is the same shape of thing as Semrush and Ahrefs. Any correction that deflates suites has to deflate it too. Doing that to both sides, the cited side turns out to be 96.7 per cent suite against the uncited side’s 88.9 per cent, so a single deflation factor applied everywhere widens the ratio rather than narrowing it, from 19.3 to 29.6 at a factor of a tenth. Deflate the two giants harder than SE Ranking, which is defensible since Semrush sells far more products than SE Ranking does, and it narrows to 8.6. The direction depends on relative factors we cannot measure, so we are not claiming one. What we can say is that the correction cuts both sides and that the cited side is 97 per cent a single suite, which is the fact the first version of this caveat left out.
• Search volume is modelled, not counted. These are estimates from clickstream-corrected data, not server logs, and two of the fifteen are contaminated in ways we could detect. There may be contamination we did not detect.
• Absence of demand is not zero demand. “Nothing measurable” means the tool reports no volume for that query, which usually means it falls below a reporting floor, not that literally nobody searched.
• Nothing causal. We changed nothing. We looked up two numbers about other people’s companies and put them beside each other.
• We are in the data. EchoWi sits in the same catalogue and was measured on the same terms as every other tool here, and we have left our own demand figure out on purpose.

Common Questions About Search Demand and AI Citation

Does brand search volume predict whether AI will cite you?

On this evidence, barely. The fifteen tools that were cited in none of our eight runs draw 151,387 United States searches a month between them, or 141,187 once two homonym-contaminated names are removed. The three cited tools with measurable demand draw 7,864. One vendor, SE Ranking, has both real demand and citations in all three markets, so the relationship is not inverted, it is weak.

Which AI visibility tools have the most search demand?

In the United States on 7 August 2026, Semrush at 83,144 a month and Ahrefs at 39,937 are an order of magnitude ahead of everything else in our verified catalogue. Neither was cited in any of the eight runs in the companion study.

Why do you exclude some search volumes?

Because the query is not the company. evertune is dominated by a guitar bridge and gives itself away with a $1.86 cost per click when every real business tool here sits between $9 and $26. profound ai is shared with a mammography product at a cost per click we cannot use to separate them. We report both and show the total with and without them.

How do you measure demand for a brand whose name is a common word?

Add the disambiguating term and compare the cost per click. omnia in Spain returns 5,335 searches at €0.34, which is the Latin word. useomnia returns 90 at €24.27, which is the company. The volume drops by a factor of fifty nine and the value per click rises by a factor of seventy one. The expensive query is the real one.

Does this mean the biggest tools are bad products?

No, and nothing here supports that. Several of the fifteen are the best funded and best known products in the category. The claim is narrower: the pages that answer a category buying question are usually not vendor pages, so being the most searched vendor does not put you in the answer.

Where can I check the tools you are counting?

Our verified catalogue lists every tool with its price and engine count read from the vendor’s own page on a stated date. The citation data is in who the AI names in three markets, with the full list of domains cited in each one.