Most of the Search Volume for AI Tools Belongs to Someone Else

We were about to review two AI visibility tools we had not covered yet. The keyword data made them look like the biggest names in the category: Quattr at 27,100 searches a month, Superlines at 9,900, both at a difficulty under 10.

That would make Quattr more searched than every tool in our comparison combined.

It is not. Disaggregated against clickstream data, “quattr” is 627 searches a month and “superlines” is 53. Ninety-eight and ninety-nine per cent of those numbers were other companies: a financial data platform, a bus operator in the Philippines, and a wholesale clothing distributor in Atlanta.

Disclosure: EchoWi competes with both tools in this article. Every figure below was pulled with the United States and English set explicitly on 6 August 2026, and prices were read from each vendor’s own page the same day.

The short version

1. Two vendor names lost 98% and 99% of their apparent demand once close variants were separated.
2. This is the third and fourth case we have documented, after “aeo” and “wellows”.
3. The cause is not a bad tool. It is how Google Ads groups close variants, which every keyword tool inherits by default.
4. The fix costs one setting, doubles the price of the query, and changes the decision completely.
5. The same failure appears inside AI visibility measurement, which is the part that should worry a buyer rather than amuse them.
6. Disaggregation is not enough. Three more names, measured on 7 August 2026, have correct disaggregated numbers that still belong to other companies. One of them owns none of its own first page.

What the numbers actually are

Same keywords, same market, same day. The only difference is whether close variants were separated.

And here is where the rest of it went.

United States and English set explicitly, 6 August 2026, clickstream disaggregation enabled.

The Quattr case is the sharpest, because the near-miss is another company in the same broad industry. quartr.com sells structured earnings-call and filings data to investment professionals. Its name differs by one letter, it has more than thirty times the search volume, and a marketer glancing at a keyword report would have no way to know they were looking at it.

Why the grouped number is wrong in a specific way

This is not a bug and it is worth understanding, because the same logic runs underneath a great deal of what this industry reports.

Most keyword volumes originate in Google Ads, which groups plurals, misspellings and close variants into a single reported figure. That is the right behaviour for an advertiser, because an ad on “superline” will also serve on “superlines” and the merged number tells them what they can actually buy.

It is the wrong behaviour for anyone asking who is searching for a company, because the merge is done on string similarity and not on meaning. “Quattr” and “quartr” are one edit apart and mean nothing to each other.

Clickstream disaggregation splits them back out using observed browsing behaviour rather than string distance. In every tool we know of it is off by default and doubles the cost of the query, so the honest description of what happened here is that the wrong number is the cheap one and the default one.

The SERP tells you before the volume does

There is a free check that would have caught all four cases, and it takes about ten seconds: search the term and read the first page.

For “superlines”, the results include the tool at position one and then a Philippine bus company at four, five and nine, a wholesale apparel business at seven, twelve and nineteen, a Los Angeles fashion district listing at ten, a water meter manufacturer at fourteen and a different AI product at seventeen. Five distinct businesses, one name.

“Quattr” behaves the opposite way and that is the interesting part. Its SERP is almost entirely the tool: its own site first, then LinkedIn, G2, Capterra, ZoomInfo and a run of review sites. Only one result on the first two pages belongs to somebody else.

So the SERP said Quattr owned its name and the volume said it had 27,100 searches, and both were misleading in different directions. The SERP was right about ownership and silent about size. The volume was right about size and silent about who. You need both, and the second one costs money.

The failure that disaggregation does not fix

Everything above is one mechanism: close variants merged on string similarity. Disaggregation fixes it, for double the money.

Three names we measured on 7 August 2026 fail a different way. The disaggregated number is correct, and it is still not theirs. Separating variants does not separate meanings, and when a company is named after a common object the meanings are what you are up against.

ZipTie is an AI search visibility tool. The string “ziptie” reports 12,089 searches a month in the United States, with variants already separated, at a difficulty of 14. Read the first page and the tool is not on it. There is a cable-tie wholesaler at one and seventeen, Amazon, Home Depot, a fastener buyer’s guide, the Wikipedia article for “Cable tie”, Merriam-Webster and Cambridge both defining “zip tie”, and a Valorant streamer using the handle. None of the thirteen organic results belong to the company. Google’s own knowledge panel for the string is the plastic fastener.

Its real demand lives on the strings nobody else wants: “ziptie ai” is 70 a month and “ziptie dev” is 50.

Omnia is an AI visibility platform. The string “omnia” reports 5,335 searches a month in Spain at a difficulty of 13. Two of the seventeen organic results are the company, its own site and its LinkedIn page. The other fifteen belong to at least fourteen different organisations: an IT infrastructure group, an insurance underwriter, a headhunting firm, a Dutch pagan folk band with its own Wikipedia article, a Seville digital agency, a Bulgari perfume, a Catalan property developer, an e-learning platform, a Catalan government social programme, a crypto token and a tarot reading service. The unambiguous string, “useomnia”, is 90 a month.

Daydream is the one that flatters itself least. “daydream ai” reports 999 searches a month in the United States. One of the sixteen organic results is the AI search company, and it turns out to be an agency with no published price rather than a tool. Thirteen belong to a fashion shopping platform with fifty million dollars of funding and a Wikipedia article. The remaining two are a desktop video editor and a sample generator for music production. Four companies, one word.

The tell is what an advertiser will pay

We expected search intent to separate these, and it did not. All three contaminated strings came back “informational”, and so did the clean one we checked against. Intent did no work at all here.

The cost per click did. Same call, same day, same settings.

$27.64 ÷ $3.31 = 8.4 €24.27 ÷ €0.34 = 71

Advertisers solved this before we did, and their bids are public. Nobody bids €24.27 for a click on a word. They bid it for a click from a person who has already decided which company they mean. The €0.34 that “omnia” clears is the price of a word that means everything; the €24.27 that “useomnia” clears is the price of a buyer.

The control is the useful part. Serpstat is a single company’s name with no everyday meaning, and it sits at $15.82 with no cheap twin underneath it. There is no second, larger, cheaper number to be fooled by, because there is no second business.

Three contaminated names and one clean one is not a law, and we are not offering it as one. It is a flag: when a brand string is much cheaper per click than the same brand plus a qualifier, the cheap number is measuring somebody else, and the SERP will tell you who in about ten seconds.

We then used it wrong, on purpose, and it failed

The next day we tried the flag as a prediction and got it backwards, which is worth writing down because the way it failed says what the test is.

Two more vendor names looked cheap against the rest of their category. Across the AI visibility tools we track, a brand click clears roughly $10 to $23: profound ai $12.31, semrush ai visibility toolkit $12.56, goodie ai $18.93, peec ai $21.84, otterly ai $22.78. Two names sat far below that line, bluefish ai at $4.18 and wellows at $0.93, so we predicted both were contaminated and went to check.

bluefish ai is clean. Every one of the seventeen organic results is the marketing platform. No fish, no text editor. The prediction was wrong.

It was wrong because we compared the name against other companies instead of against itself, which is not the test. Run the real one and it fires hard:

$4.18 ÷ $0.33 = 12.7

The bare name carries 26,655 searches at a third of a dollar a click, and it is not the company at all. Nothing about the category average told us that; the within-name comparison told us immediately. The flag compares a name to itself with a qualifier attached. Comparing it to its neighbours is a different idea and it does not work.

And a second signature, from the one that was contaminated

wellows did turn out to be mixed: five of seventeen results belong to a compression-sock brand spelled wellow, singular. But its gap is small, $0.93 against $1.95 for wellows ai, so the ratio test barely registers it.

What gives it away is something else entirely:

The same price, to the cent, as another company’s brand term. Two strings clearing an identical cost per click are being auctioned to one audience, and when one of them is a larger business with a nearly identical name, the smaller one is buying its neighbour’s traffic. That is a second signature and it costs the same nothing to check.

The two tools, since we looked them up anyway

Neither is a bad product, and neither did anything wrong here. The measurement failure is ours and the industry’s.

Superlines is built by Grew Oy, a Finnish company, and publishes its prices, which puts it ahead of most of this market before you look at anything else.

Re-read from Superlines’ own pricing page on 8 August 2026. Every tier carries a seven-day free trial, and a parallel agency ladder runs at the same three prices, swapping brands for client brands.

The thing worth flagging. The engine count is a pick, not a total, and it never moves: Growth costs 4.8 times Starter and still selects three engines, so what you buy going up is prompt volume rather than coverage. An earlier version of this table said the agency tier cost €299 and included five engines; it costs the same as the brand tier and the five is client brands. Corrected on 8 August 2026, and the review carries the full correction.

Quattr Inc publishes no prices at all. Its pricing page routes to a demo, so we cannot tell you what it costs and neither can anyone else outside a sales call. It names Google AI Overviews and AI Mode, ChatGPT, Perplexity, Gemini and Claude, and it positions itself across SEO, AEO and GEO rather than as a pure visibility tracker. It also publishes an unusually large set of free tools, which is a real thing to like: like Semrush’s free checker, it lets you form an opinion before a call.

Hidden pricing is now the majority position in this category. Bluefish, Brandlight, Peec AI and Evertune all do the same thing, and we have said the same thing about each of them: it is defensible for enterprise sales and it makes a comparison article impossible to write honestly.

The part that should actually worry you

Everything above is a keyword research problem, which is annoying. The same failure inside AI visibility measurement is expensive.

If your brand name is a common word, a measurement that counts mentions is counting other people. A tool reporting that you were named in 40% of answers cannot tell you whether the model meant your company, and almost none of them publish how they resolve it. This is entity disambiguation, and it is the least glamorous item in our glossary and the one that decides whether a number means anything.

We have watched it happen in our own category twice. “AEO” is the acronym this whole field uses for answer engine optimization and it is also the ticker for American Eagle Outfitters, which is why we refused to build on the American volume and wrote the Spanish version separately after measuring that the Spanish SERP splits the other way.

The practical version, for a brand:

• If your name is a dictionary word, an acronym or one letter from a bigger company, assume every off-the-shelf number about you is contaminated until the vendor shows you otherwise.
• Ask any tool you are evaluating how it decides that a mention refers to you. “We look for the brand name” is not an answer, it is a description of the problem.
• Ask to see the raw answer text, not just the count. If you cannot get back to the sentence, you cannot audit the number.

And for a founder naming a company: a name one edit away from a larger business is a permanent tax on every measurement you will ever run, on your own marketing and on anyone trying to write about you fairly.

How to check this in your own reporting

Step five is not padding. Our three-market study found that the stable answer sets in the United States, Spain and France did not overlap at all, and the AEO case splits by country in exactly this way.

What this does not show

• Nothing about either product’s quality. We have not tested Quattr or Superlines, and a name collision says nothing about the software behind it.
• We did not disaggregate every keyword we have ever used. Seven names, each checked because something looked wrong. There are almost certainly more.
• The disaggregated figures are estimates too. Clickstream data is a sample of observed browsing, not a census, and it carries its own error. It is better than the grouped number, not exact.
• The cost-per-click pattern is three cases and one control. That is a flag worth checking, not a rule, and we have not tested it against a set large enough to call it one.
• This is mostly one market. Every figure is United States and English except “omnia”, which is Spain and Spanish and is labelled as such. We have not run the same check in France for any of these names.
• We cannot show that fixing this changes anyone’s outcome, because that would need a controlled test and no one has published one. A July 2026 review of 45 GEO studies found no technique with a demonstrated causal effect, and this is a measurement point rather than a technique.

Common Questions About This Study

Why was the search volume wrong?

It was not wrong, it was answering a different question. Google Ads groups close variants such as plurals and misspellings into one reported figure, which is correct for an advertiser buying that group and misleading for anyone asking how many people search for a specific company. “Quattr” was merged with “quartr”, “quattro” and “quatre”, which between them are more than thirty times its real volume.

How do I get the real number?

Turn on close-variant or clickstream disaggregation in whichever keyword tool you use. It is off by default in every tool we know of and it roughly doubles the cost of the query, which is why the wrong number is also the cheap one. Then sanity-check it against the search results page, which is free.

How much did the numbers move?

“Quattr” went from 27,100 a month to 627, so 97.7% of the apparent demand was other companies. “Superlines” went from 9,900 to 53, which is 99.5%. “Wellows”, which we checked earlier this year, went from 5,400 to 624. All three were measured with the United States and English set explicitly.

Does this affect AI visibility measurement too?

Yes, and more expensively. A tool counting how often an assistant names your brand has to decide that a mention refers to you rather than to a company with a similar name, and almost none of them publish how they do it. If your brand name is a common word or a near-miss for a larger business, treat every count as contaminated until the vendor can show you the answer text behind it.

What are Quattr and Superlines, then?

Both are real AI visibility platforms. Superlines is built by Grew Oy in Finland and publishes prices from €79 a month for 50 prompts across three chosen engines, with a seven-day trial. Quattr Inc positions across SEO, AEO and GEO, names Google AI Overviews and AI Mode, ChatGPT, Perplexity, Gemini and Claude, publishes a large set of free tools, and publishes no pricing at all.

Does clickstream disaggregation always fix it?

No, and that is the second half of this study. Disaggregation separates close variants, not meanings. “ziptie” reports 12,089 searches a month in the United States with variants already separated, and none of the thirteen organic results on its first page belong to the AI search tool of that name. “omnia” reports 5,335 in Spain and two of seventeen results are the company. When a business is named after an everyday object or a Latin word, the number is right and the ownership is still wrong.

How can I spot a contaminated brand keyword quickly?

Compare what an advertiser will pay for the bare name against the name plus a qualifier, which costs nothing extra because both come back in the same call. In our cases the gap was large and one-directional: “ziptie” clears $3.31 a click and “ziptie ai” clears $27.64, and “omnia” clears €0.34 against €24.27 for “useomnia”. A name with no everyday meaning, like “serpstat” at $15.82, has no cheap twin at all. Then read the first page of results to confirm who owns it. Search intent did not help us here: every string we checked came back informational, contaminated or not.

Should I rename my company if it collides?

Probably not on this basis alone, because renaming costs far more than the measurement error. But it is worth knowing the tax you are paying: every keyword report, every share-of-voice number and every AI mention count will need extra work to be trustworthy, and journalists and analysts writing about you will hit the same wall. If you are still choosing a name, check the search results page for it first. It takes ten seconds and it is the cheapest due diligence available.