LiamVi

← All articles  ·  17 September 2026

A wide card with six identical tall windows cut into its face in one row. Five of the windows are empty; the third from the left is filled solid orange. Marked One list above The others.

Choosing an AI visibility tool in 2026: how much the recommended lists agree, and who wrote them

Six of the nine pages Google returned for ai visibility tool on 14 September 2026 answer the question with a recommended list, and across those six lists we counted 32 distinct products — of which 2 appear on all six and 21 appear on exactly one. Nobody on that results page reports that. Every AI visibility tool list we opened tells you which products to look at; not a single page tells you how much its list agrees with the list in your other tab.

There is a second thing none of them tells you, and it is the one to carry into any ranking you read: 4 of these 6 lists were written by a publisher that sells a product on the list, and in all 4 that product is ranked first. Who wrote each list, and what each says about how it decided, is the second half of this article.

The first list we have to account for is our own. Our page on this keyword, published on 10 August 2026, says: "Three verifiable products track AI visibility today: Semrush's AI Visibility Index, Profound, and LiamVi." There is no census behind that sentence, no method, no date on the count — and we put ourselves on our own list of three. Of the 32 products the six lists name between them, that page named two, and added a third that appears on none of them: liamvi.com is on none of the six lists we opened.

This article is a count of how much six published lists agree, made on a single day from a single results page. It is not a ranking, not a test of any product, and not a recommendation. We have used none of these tools, we are on none of these lists, and nothing here says that a product named six times does its job better than a product named once.

Something else belongs at the top rather than buried later: somebody has already done a version of this work. Ryan Doser's page of 1 September 2026 checks 22 roundups and reports which LLM tracking product survives the ones written by publishers with no tool to sell. It does not appear on this keyword's results page — which is why reading the results would never have found it, and why every claim of absence below is scoped to the 9 pages Google returned for this keyword, never to the whole web.

If what you want first is the vocabulary — what a brand mention is, how it differs from a citation, what share of voice counts, what any of these products actually track across ChatGPT, Gemini and the rest of the AI search surfaces — our earlier page on this keyword sets those definitions out and this article does not repeat them. What follows assumes you have decided you want AI search visibility tracking and are trying to work out whose recommendation to use.

How much do the recommended lists actually agree?

The six recommended lists agree on 2 products out of the 32 they name between them: Peec AI and Profound. Semrush is named on 5 of the 6 lists and AthenaHQ on 4. Eleven of the 32 products appear on two lists or more. The remaining 21 of the 32 appear on exactly one list and nowhere else in the set, which is the more useful half of the finding: for most of the products being recommended to you as an AI visibility tool, you are reading one publisher's judgement and no second opinion exists anywhere on that results page.

One upright plate filled solid orange, with two identical navy tabs standing side by side on its top edge, entirely above the orange face. Marked Two names above the tabs and One product beneath the plate.
Names are counted as each page writes them, with five merges. The merge that two of the lists disguise between them is settled only by the vendor's own site, which brands the product one way and lists the second name as a feature of the same platform. Read on 14 September 2026.

Here are the six lists, whole, in the order each page ranks them, so you can re-count them yourself and disagree with us.

| Publisher | Date shown on the page | Author shown | The list, in the page's own order | |---|---|---|---| | Zapier | none shown | none shown | Profound · Otterly.AI · Peec AI · ZipTie · Similarweb · Semrush · Ahrefs · Clearscope | | Frase | 18 Feb 2026, updated Aug 2026 | none shown | Frase · Profound · Peec AI · Semrush AI Toolkit · Athena · SE Ranking · Positive Surfer · Writesonic · Otterly.ai · ZipTie | | Profound | 30 Jan 2026 | Nick Lafferty | Profound · Scrunch · Otterly · AthenaHQ · Rankability · Peec AI · Semrush · Ahrefs · AirOps · Writesonic · Conductor · BrightEdge · Bluefish · Evertune · Adobe LLM Optimizer · SE Visible · AIclicks · Brandlight | | Marketer Milk | 23 Dec 2025 | Omid Ghiam | AI Product Rankings · Peec AI · Profound · Scrunch AI · Hall · Nimt.ai · AthenaHQ · Mentions · BrandLight · Google Analytics 4 | | Backlinko | 19 Jun 2026 | Leigh McKenzie | Semrush · Profound · ZipTie.Dev · Peec AI · Gumshoe.AI | | Analyze AI | 20 May 2026 | "Ernest" | Analyze AI · Profound · AthenaHQ · Peec AI · Ahrefs (Brand Radar) · Semrush (via Otterly) · Surfer (AI Tracker) · Wix |

We counted 59 entries across those six lists, resolving to 32 distinct products.

The method is short enough to repeat. We took the 9 results Google returned for this keyword on 14 September 2026, from a capture with `location: United States`, and opened every page. Six carried a list of AI search visibility products; three did not. For each list we recorded every numbered or headed entry — a product the page puts forward as a pick, not a tool mentioned in passing — in the order the page ranks it. Then we counted how many lists each product appeared on. A browser and a notepad are the whole toolkit; we used no data of our own and nothing you cannot open yourself.

Counting names means deciding when two names are the same product, so here is the rule we used, and it is the part worth checking hardest. Names are counted as each page writes them, with five merges: Otterly.AI, Otterly and Otterly.ai; ZipTie and ZipTie.Dev; Athena and AthenaHQ; Scrunch and Scrunch AI; and Positive Surfer, on Frase's list, with Surfer (AI Tracker), on Analyze AI's. That fifth merge is one two lists disguise between them, and only the vendor's own source settles it: Surfer's site now brands the product Positive Surfer inside the Positive portfolio, and lists AI Tracker under its platform as a Surfer feature that tracks "how your brand appears in AI tools like ChatGPT." Two names, one product, two lists.

Two smaller rules finish the job. A vendor's named suite counts under the vendor, so Semrush AI Toolkit is not a second product here. And an entry naming two companies at once counts once, under the name heading it — which is how Semrush (via Otterly) lands under the first of its two names. That rule has a cost worth knowing, because it is this article's own lesson turned on itself: the name heading an entry is not always the company that built the product. Ahrefs (Brand Radar) is Ahrefs' own product. Otterly is not. Semrush's documentation calls it "an app by Otterly.ai", and Otterly's announcement describes "our partnership with Semrush and the launch of our AI Search Monitoring App in the Semrush App Center" — distribution, not an acquisition. We infer no ownership from a parenthesis, and neither should you: open the source page of whichever vendor the entry names.

Change those merges and the 32 moves by a product or two. The shape does not move: most named products appear on a single list.

Four limits sit on that count and they belong here rather than in a closing caveat.

Three of the 9 results carry no list at all. Semrush's AI SEO overview, Ahrefs' AI visibility checker and Rankscale's homepage are vendor pages about a single product. Opened today, the first of those carries no article at all — navigation, headings and Semrush's own product names — which makes it a landing page rather than a piece of writing, and we do not quote it as one. Those three are not counted as agreement or disagreement; they are not lists. It is worth noticing anyway: 3 of the 9 pages Google hands you for this keyword are a vendor selling the thing you are researching.

This count comes from a single results page, captured on a single day, in a single location. A results page is a photograph of a moment. Run the same search next month and the set of pages will have moved, and so may the count.

Agreement is not quality. A product on a single list may be the best tool in the category; a product on all six may be the best-marketed. We have tested none of them, and this count says nothing about what any product does.

The sample is small and early. Six lists is six lists. We will run this count again on a later capture and publish what changed, including a result that contradicts this one.

Part of the disagreement has a plain explanation, and it is worth separating from the rest: these are not six lists of the same kind of thing. The 18-entry list carries enterprise suites — Adobe LLM Optimizer, Conductor, BrightEdge — that the 5-entry list does not include, and Backlinko's author describes testing "across enterprise solutions and newer entrants" before picking five. A page writing for an agency with an enterprise budget and a page writing for a solo marketer tracking a single brand will not name the same AI search monitoring products, and neither is wrong to. What none of them does is tell you which kind of list you are reading, or which kind of buyer it was written for.

Doser's page, which we found by searching rather than by reading the results, cuts across all of this usefully. He checked 22 roundups and found that seven were published by people with no tool to sell in the category. Across those seven he reports Peec AI as "the only tool named in all seven independent roundups", with Profound and Semrush each appearing on six of the seven. His frame is a set of roundups he selected; ours is the set this keyword returns. They are different samples answering slightly different questions, and they land on the same product. He also states plainly that he sells a competing product and left it off his own list, and that a tool on his list pays him a commission — a disclosure more explicit than most of the pages we counted, and worth saying so.

Weigh this against everything above: we build an AI search visibility tool ourselves, and we are on none of these lists. That is a reason to check our count, not a reason to trust it more. The lists are printed above precisely so you can.

Who wrote the list you are reading?

The list you are reading was, in 4 of the 6 cases here, written by a publisher that sells a product on it, ranked alongside its own competitors — and in all 4 that product is ranked first. That is a structural fact about how buying guides in AI search get written.

Six identical upright plates standing in one row, four on the left and two on the right across a wide gap, each carrying one short bar at its head. One orange rule lies beneath the four, marked Own product first; the two are marked No house product.
Four of the six lists on this results page were published by a company that sells a product on the list, and in all four that product is ranked first. The other two lead with somebody else's. Read on 14 September 2026; a later capture may return a different set of pages.

Frase's list of ten opens with Frase. The list of eighteen published on Profound's own blog opens with its own platform, described there as "an agentic marketing platform that shows you where your brand is missing from AI answers". Analyze AI's list of eight opens with Analyze AI.

Backlinko's list of five opens with Semrush, and the page itself tells you why: "Backlinko is owned by Semrush." The same page carries a reader-offer block — "Backlinko readers get a 14-day trial for premium features" — in plain sight. Both are disclosures, printed by the publisher. The right response to a disclosure is to credit it: a page that tells you who owns it has handed you what you need to read it properly.

The two lists that do not lead with a house product come from Zapier, a publisher that sells nothing in this category, and Marketer Milk, which carries its own disclosure: "Some of the links in this article may be affiliate links, which can provide compensation to Marketer Milk at no cost to you if you decide to purchase a paid plan."

Every one of the six tells you something about how it decided. What separates them is whether anybody says they used the products, and on that line the six split 4 to 2.

Four say they tested. Zapier's author writes: "I've tested over 20 LLM monitoring tools in this developing market", and describes "creating accounts, watching their demo videos, booking product walkthroughs, and reading the product documentation." Analyze AI writes: "We spent 90+ days testing every tool on this list across real brand projects. We tracked the same prompts, compared outputs, and documented what each platform can and cannot do." Marketer Milk's author writes: "I've spent the past 6 months just testing as many of these tools as I can", shows screenshots of himself inside most of them, and names the one he did not use — "the only platform I didn't actually use myself on this list." Backlinko's author writes: "I've tested more than a dozen LLM tracking platforms across enterprise solutions and newer entrants."

The other two state criteria and no test. Frase hands the reader six of them — "Use these six criteria when you evaluate any platform on this list" — and says "Two things decide the order." The 18-entry list sets out what to look for in a tool for agencies and rates each product on where it wins and where it falls short. Neither page says anyone used what it ranked. That difference is worth naming precisely, because it is not the difference between a method and none: criteria tell you what somebody thinks matters, a testing claim says somebody did the work, and only the second is something you can hold them to.

None of this means any list is dishonest. A vendor writing about its own category often knows it better than anybody, and putting your own product first on your own blog is not a deception — it is what a blog is. There is no evidence here that any writer acted in bad faith, and no product is bad because of where it sits on somebody's page. This article makes neither claim. What the pattern changes is how you read the ranking: the order on a vendor's page is a marketing decision as well as an editorial judgement, and the page will usually not separate the two for you.

So the practical version is small. Read the byline and the ownership line before the ranking. Then check whether the page says anyone used what it ranked. A list that says somebody used the products can be argued with; a list that gives you only its criteria is asking you to take its shortlist on taste — and on this results page, 2 of the 6 carry no byline at all: Zapier, which shows neither an author nor a date, and Frase, which shows a date but names nobody.

What did our own page on this keyword get wrong?

Our own page got 3 things wrong, and the first is the one this count exists to correct: it asserted that 3 products track AI visibility and put LiamVi among them. A correction is an article rather than a quiet edit, so here are all three, in the order they matter.

The list of three was an assertion. That page says "Three verifiable products track AI visibility today" and names them. There is no method behind the number three, no date on the count, and nothing a reader could verify. Two of its three products turn out to be among the four most-named in this census, so the instinct was not wrong. The form was. We published a count of a market without counting anything — and we were on our own list.

That page also treated the category as though it had three members. Six published lists name 32 between them, and a reader deciding what to use is choosing inside that whole set rather than inside ours.

It printed rates without their denominators. The same page carries the sentence "cited and ignored pages answered searcher questions at almost identical rates (58.8% against 58.2% in our sample)", with no sample size anywhere in it. Our rule is that a rate never travels without its denominator and that small results publish as raw fractions, because a fraction is checkable and a lone percentage reads bigger than the work behind it. That page went live on 10 August 2026; the rule was written on 12 August 2026, two days later. That is the honest sequence, and it does not make the sentence right. The measurement belongs to the pages that publish it with its sample attached, and we are not repeating it here.

Its sources were checked on 4 August 2026 — 41 days before the capture behind this article. In a category where the results page turns over this fast, a source date is part of the claim.

Then the part that is least comfortable. liamvi.com appears on none of the six lists, and our own page is the only place in this census where LiamVi is named at all.

What this page does instead is narrow and checkable: where the earlier version asserted which products exist, this version counts what six published lists actually say, prints them, dates them, and names who wrote each.

How can I check my AI visibility for free?

You can check your AI visibility for free in 2 ways: run a vendor's free checker against your domain, or ask the AI search assistants yourself and record what they name. They are worth different things, and knowing which is which matters more than the number either produces.

A large upright rectangle outlined in navy, empty inside, with one small solid orange rectangle standing wholly inside it near its upper left and touching no side. Marked Limited preview beneath the small rectangle and The paid tool beneath the large one.
One vendor's free checker returns what its own page calls a limited preview; the paid product tracks more. The figure shows containment, not a measured share — we have measured neither side, and where the line falls is a commercial decision rather than a measurement one. Read on 14 September 2026.

This is one of the four People Also Ask questions Google returned for this keyword on 14 September 2026, in the same capture the census came from — and counting how much six lists agree cannot answer it. So here is what can, and where each answer stops.

The vendor free checks are the fastest. Ahrefs' AI visibility checker states "No signup required. Results in seconds" and returns what the page calls "a limited preview" — total AI mentions, mentions by AI platform, top five topics, top five cited domains and top five cited pages. Frase's page offers a free checker that "shows who AI cites for your domain", and says of its free tools: "Neither needs a card."

Each of those is run by a company selling the paid product. That is not an accusation — it is why a free tier exists at all, and it is the ordinary shape of a free trial in software. It does mean the line between what the free check shows you and what it holds back for the paid AI visibility tool is a commercial decision rather than a measurement one. Read the free number as a sample of what the product tracks, not as the measurement itself.

The manual check needs nothing from anybody. Write down the handful of prompts your buyers would actually use — not your keywords, the questions — and put each to ChatGPT, Gemini, Perplexity, Claude and Copilot in turn. Record which brands get a mention in the answer and which pages get cited as sources, and write down every competitor named beside you. Keep mentions and citations apart while you do it: a mention means the assistant named your brand, a citation means it linked your page, and AI search monitoring products bill for tracking both. Do it twice, on different days. That is the same brand visibility tracking a monitoring platform automates, done by hand on a sample small enough to read in an afternoon.

Then the limit, because it applies to the free check, the paid platform and the manual version equally. Whatever number you end up with is a count over a prompt list somebody chose. As our page on AI visibility analytics puts it: "Change the prompt list and the number changes, with no change at all to your website." Add ten prompts from a market where you are strong and your share of voice rises; add ten from a market you have never written for and it falls. Nothing about your site moved.

The same applies across time. Ask the same assistant the same question twice and the answer, and its citations, can change — that is our own measurement, from a 33-keyword citation sample collected mostly from ChatGPT, and it is written up with its limits here. It is why a single free check is a reading rather than a verdict. Two honest tracking products sampling at different moments will disagree, and the useful question to put to any vendor is not is your data accurate but what is your prompt list, how often do you sample, and how much do your own numbers move between runs.

Here is what to do with that tomorrow. Run a free check, write down the prompt list it used or the list you chose, run it again a week later, and see how much moved. If the answer is a lot, you have learned something real about the measurement before paying for it, and you can make that a monthly step in your own workflow. Whether a paid platform is worth it after that depends on how many prompts, markets and competitor brands you need tracked, and that is a judgement about your business no count of ours can make. If the question you arrived with is whether to use separate products for AI search visibility and for ordinary SEO, we have written that up in AI visibility platforms with SEO capabilities.

What this does not tell you

This count does not tell you whether any of these 32 products works. We tested none of them, the count is not a measure of what any of them does, and agreement between lists is not evidence of quality.

Thirty-two is not a count of the market. It is what six pages named on a single day. Other visibility tools exist that none of the six gave a mention, ours among them, and new ones keep arriving.

A reader searching from elsewhere may be handed a different set of pages entirely, and the count would have to be run again for it.

The absence claim is scoped. When we say no page reports how much its list agrees with another, we mean the nine pages Google returned for this keyword on that date, plus a single free web search. A search that finds little is not proof that little exists — Doser's page is the live demonstration of exactly that, and we found it only because we went looking.

Our own instrument has been wrong in public before. We tested our fitted content score against blind quality reviews run by independent reviewer agents, and it pointed the wrong way, at rho −0.61 across the nine articles it covered. That write-up is in our piece on content optimization tools, and it is the honest answer to anybody asking whether a vendor's score can be trusted, ours included. It is also the standard we are asking you to hold this count to: any visibility tool's number, including ours, is an instrument somebody built and should be checked the same way.

The receipt is owed. Our research page carries our samples and our limits, and it does not yet carry this census. It should, and it will.

We keep measuring. This count will be re-run on a later capture, and we will publish what changes.

What people actually ask us

These 4 questions are the ones Google returned as People Also Ask for this keyword on 14 September 2026, in the same capture the census came from, and each answer below stands on its own.

What is the best AI visibility tool? We are not going to name one, and the reason sits in this article: we have tested none of them, we are on none of these lists, and the six lists that do answer the question agree on 2 products out of the 32 they name between them. If you want the shortest honest summary of where the published consensus sits, Peec AI and Profound were the only products named on all 6 lists we opened on 14 September 2026 — but that is a report of what six pages said, not a verdict on what you should buy.

What are AI visibility services? The phrase usually separates buying software from buying somebody's time: a tool hands you the monitoring data and leaves the work to you, while a service is an agency or consultant who runs the prompt tracking, reads what ChatGPT and the other assistants return, and often does the content optimization on your behalf. We did not survey that market, this count measured none of it, and we are naming no provider — the questions in this article about method and disclosure transfer to a service, but our numbers do not.

Does a product appearing on more lists mean it is better? No. It means more publishers included it, which can follow from age, marketing reach, an affiliate programme, or genuinely being good — and none of those is distinguishable from outside. The count is more useful in the opposite direction: when a list is the only place a product gets a mention, you know you are reading a single publisher's judgement rather than a consensus, and you can go and use the product's own documentation instead of the ranking.

Where does your count come from, and can I check it? From the 9 results Google returned for ai visibility tool on 14 September 2026 in a `United States` capture, every page opened by hand, with each numbered or headed entry recorded in the order that page ranks it. The merge rule is in this article, the six lists are printed in full above, and every page is linked. Two of the merges are ones the pages themselves disguise, so the vendors' own source pages are linked on the claims that settle them. You can run the whole audit yourself in an afternoon with a browser, and if your capture differs from ours, yours is the more recent photograph and the one to trust.

Sources