Guides · Guide · Published 2026-09-20
AI visibility tools: what each vendor publishes about its own method
A comparison of seven AI visibility tools on four checkable criteria — engines covered, sampling method, repeats and interval, and public pricing — built only from what each vendor publishes on its own pages. We sell one of them, so there is no ranking here.
AI visibility tools are subscription products that send buyer questions to AI assistants such as ChatGPT, Perplexity and Gemini on a schedule, then record whether a brand is named and which sites the answer cites. They differ less in features than in which engines they query, how many times, and how much of that method they publish.
We sell one of the things on this page
RedClaw Labs runs a paid AI visibility measurement service. That makes any ranking from us worthless, so this page does not contain one. There is no "best", no score, no winner, and the rows are in alphabetical order, which puts us fifth out of seven with no say in the matter.
Here is how we tried to keep it honest, and how you can check us:
- Every cell comes from the vendor's own public pages. Not from a roundup, not from a review site, not from memory. Each source URL and the date we read it is listed at the bottom. Nothing here is second-hand.
- Where a vendor does not publish something, the cell says "not published". It never says "unknown, probably no". A vendor may well answer the question on a sales call. We can only report what is on the public site.
- No judgements about anyone else's product. We state what a vendor says about itself. Whether it is good is not something we are in a position to tell you.
- Our row uses the same four columns as everyone else's. We did not add a column we would win on. We come out worst in the table on engine coverage, and we are one of only two rows with no public price.
If a cell is wrong or out of date, the fix is to open the source URL and read it. That is the whole reason the URLs are there.
Why these four columns, and not a feature list
Feature lists are where comparisons of this category go to die. Every vendor has dashboards, alerts, competitor views and an export. Counting features tells you which marketing team was busiest, not which number you can defend in a meeting when someone asks where it came from.
In AI brand monitoring we set out ten things a tool in this category has to do, and deliberately named no products. Four of those ten can be checked by a stranger with a browser, before any trial, login or sales call. Those four are the columns here. The other six need an account to verify, so putting them in a public table would mean taking somebody's word for it.
Which engines it says it covers. The first question is not how good the tool is, it is whether it looks at the surface you care about. An answer of "AI search" is not an answer. This column also exposes a detail worth catching: on some products the engine list on the homepage is longer than the engine list included in the plan you would actually buy.
Whether the sampling method is published. Where do the prompts come from, how often are they run, and how does the query reach the engine? A number you cannot reproduce is a number you cannot argue with.
Whether repeats per prompt and an interval are reported. This is the column that matters most and the one nearly nobody fills. Ask the same question twice and you get two different answers, so a visibility percentage is a proportion drawn from a sample. Without the number of trials behind it, you cannot tell a real change from the sample being small. Our own worked example below shows that happening within fifteen minutes.
Whether pricing is public. Not a quality signal, just a practical one. It decides whether you can shortlist without booking a call.
One thing no column here measures is whether the tool's numbers are accurate, because that cannot be established from a vendor's website by anyone, including us. The last section is about how to test that yourself.
The comparison
All cells read on 2026-09-20 from the vendor pages listed under Sources. Alphabetical order. "Not published" means we could not find it on that vendor's own public pages on that date.
| Product | Engines the vendor says it covers | Is the sampling method published? | Repeats per prompt, and an interval? | Is pricing public? |
|---|---|---|---|---|
| Ahrefs Brand Radar | AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot | Partly. Prompts are drawn from its keyword database and expanded via People Also Ask and "semantic fanout"; custom prompts are "checked daily on every selected platform"; index figures run on a 90-day window. How the query reaches the engine is not published | Not published | Yes. Custom Prompts from US$50/month; AI Visibility Index US$199/month |
| Frase | ChatGPT, Perplexity, Claude, Gemini, Google AI. The pricing page ties engines to tier: ChatGPT and Google AI on Starter and Professional, Claude and Gemini added on Scale | Partly. "Frase tracks your prompts daily"; tracked prompts are capped per tier at 50, 200 and 500. Prompt sourcing and transport not published | Not published | Yes. US$49, US$129 and US$299 per month; US$39, US$103 and US$239 billed yearly |
| Otterly.ai | Pricing page: ChatGPT, Google AI Overviews, Perplexity and Microsoft Copilot on every plan, with Claude, Google AI Mode and Gemini sold as add-ons. The homepage names all seven | Partly. The help centre states prompts are "automatically monitored daily across all AI search engines available in your account". Prompt sourcing and transport not published | Not published | Yes. US$29, US$189 and US$489 per month; Enterprise custom, "Starting from $1,000/month" |
| Peec AI | Default: ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Gemini, Microsoft Copilot. Further engines, including Claude, DeepSeek, Grok and Mistral, are paid upgrades | Partly. The docs say an accepted prompt joins "the regular 24-hour cycle"; tracking frequency is daily on the brand tiers | Not published | Yes, on its AI instructions page: US$95, US$245 and US$495 per month for brand plans. The /pricing page rendered tier names without amounts when we read it |
| Profound | Enterprise plan: ChatGPT, Perplexity, Google AI Mode, Gemini, Microsoft Copilot, DeepSeek, Claude, Google AI Overviews, Exa Search. The free trial is ChatGPT only | Partly. The help centre states "Prompts are the queries Profound sends to answer engines daily" | Not published | No. Trial is free, Enterprise is "Custom" |
| RedClaw Labs | ChatGPT, Gemini, Perplexity, official APIs only. Google AI Overviews and AI Mode are out of scope and we say so on the product page. Shortest engine list in this table | Yes. The methodology page publishes the question generator's rules, the gate that rejects a question set, the API endpoint per engine and the pinned model identifier per engine | Yes. 3 repeats per question per engine, weekly, on paid monitoring; 1 on the free check. Every mention rate is published with a 95% Wilson interval | No. The product page says pricing is being finalised and online payment is not open |
| Semrush AI Visibility Toolkit | ChatGPT, Google AI, Gemini, Perplexity | Partly. "25 custom prompts (daily AI rankings)", with "Daily, weekly, and monthly data updates"; the knowledge base states "Prompt tracking: updated daily" | Not published | Yes. US$99/month per domain billed annually |
Reading the "not published" column
Six of the seven rows say "not published" for repeats and interval, and that is the single most consistent finding on this page. It does not mean those products run each prompt once. It means the number of runs behind a percentage is not something a buyer can establish before signing up.
It is worth being precise about what is and is not missing there. Every vendor in the table publishes a frequency: daily, in all six cases. Frequency and repeats are different quantities. Running a prompt once a day for thirty days gives you thirty daily figures, each of which is a single answer, and a day-to-day move of a hundred percentage points on a single prompt is what a single answer looks like when it varies. Three runs on one day gives you one figure with a denominator of three. Which of these a dashboard is showing you changes how much weight a week-over-week line can carry, and that is the question to take into a demo.
Where our own row loses
Engine coverage. Ours is the shortest list in the table, and every other product claims more engines than we query. Four of the six name Google AI Overviews explicitly: Ahrefs Brand Radar, Otterly.ai, Peec AI and Profound. The other two, Frase and Semrush, name "Google AI" without saying which surface that means, which is the kind of thing to pin down before you buy. We have not verified any of these claims, only recorded them. If AI Overviews is the surface you need measured, we are not your supplier, and we would rather write that here than have you discover it after paying. The same limit is stated on our AI visibility tracking page and in our methodology.
Public pricing. We are one of two rows with no price on the page. In our case payments are still being set up, so today the only way to start is to email us. That is a worse buying experience than clicking a plan, and it belongs in the table rather than in a footnote.
Cadence. Weekly, against daily elsewhere. The trade is deliberate. Three repeats on three engines once a week costs about the same in API calls as one repeat once a day, and we would rather have a denominator than a daily line. It is still a trade, and if you want to see a movement the morning after a launch, a daily product will show you something we will not.
How to check a tool yourself
Nothing above tells you whether a tool's numbers are right. Here is a test that costs an afternoon and does not require our product or anyone else's.
Take one brand. Ask the tool for its number and, critically, for the prompts behind it. Then put the same prompts to the same engine yourself, more than once, and compare. If the tool will not show you the prompts, the test is over and you have your answer.
The part people get wrong is expecting the two figures to match. They will not, and a mismatch is not proof the tool is broken. Here is what re-running the same thing actually looks like.
On 2026-09-20 we ran our free-check pipeline three times against ChatGPT through the official API, with the same three generated questions each time, for Miro in online whiteboard and visual collaboration tools, US market. The questions, verbatim and identical across all three runs, with the brand name nowhere in them:
- "Can you recommend online whiteboard tools that are best for remote teams in the US with a budget under $20 per user per month?"
- "Which visual collaboration platforms offer the best integration with popular project management software for small businesses?"
- "What are the top online whiteboard vendors known for strong security features suitable for healthcare organizations?"
| Run | Time (UTC) | Mentions | Point estimate | Wilson 95% |
|---|---|---|---|---|
| 1 | 07:24 | 3 of 3 | 100% | 43.8% to 100% |
| 2 | 07:29 | 3 of 3 | 100% | 43.8% to 100% |
| 3 | 07:39 | 2 of 3 | 66.7% | 20.8% to 93.9% |
| Pooled | 8 of 9 | 88.9% | 56.5% to 98.0% |
Fifteen minutes apart, same engine, same questions, and the headline figure moved by more than thirty points. The intervals on runs 2 and 3 overlap across most of their range, so that move is not evidence that anything changed. It is what a sample of three looks like. The arithmetic is on the methodology page, and you can reproduce it in a spreadsheet.
Two more things fell out of the same three runs that are worth knowing before you audit anybody:
The competitor list moved too. Run 1 named Lucidspark, Mural, FigJam, Figma and Stormboard. Run 3 named Lucidspark, FigJam, Figma, Microsoft 365 and MURAL. Same question set, same quarter of an hour, a different set of names. A competitor chart built on one run is a snapshot of one answer.
"Mural" and "MURAL" came back as separate strings across the runs. Whether a tool treats those as one brand or two is an alias-matching decision nobody advertises, and it silently changes every competitor count on the dashboard. Ask about it.
So the test is not "does the tool's number equal mine". It is:
- Ask the vendor how many runs sit behind the figure on screen. If the answer is one, the figure is an answer, not a rate.
- Re-run the prompts yourself at least three times and see whether your own spread brackets the tool's number. If it does, the tool is consistent with what you measured.
- Check whether a change the tool has flagged as a movement is larger than the spread you just produced by doing nothing.
That last check is the one that pays for the afternoon. Full raw output for these three runs is kept in our article run records alongside the numbers above, and the ongoing version of this measurement is what our tracking service does weekly, across three engines, with the repeats stated on the report.
Questions
What are AEO tools, and are they different from AI visibility tools? In practice the phrases are used for the same category. AEO stands for answer engine optimization, and vendors using that label sell the same thing: a product that puts questions to AI assistants and records what the answers say about you. The label tells you about the vendor's positioning. The four columns above tell you what the product does.
What is the best ChatGPT SEO tracking tool? We are not going to answer that, because we sell a product in the category and any answer we gave would be marketing. What we can offer is the order to decide in. Start with which engines you need, because that removes most of the list immediately. Then ask each survivor how many runs sit behind a percentage. Then look at price. A tool that covers your surface and tells you its denominator beats a tool with more dashboards.
Is an LLM SEO tool the same as an AI visibility tool? The names overlap, but products under the "LLM SEO tool" label often bundle two different jobs: measuring where you stand, and producing or rewriting content to change it. Those are separable purchases, and the measurement half is the only half this page compares. A supplier who both grades your visibility and sells you the content meant to fix it is marking its own homework, which is worth knowing whoever you buy from. We only sell the measurement.
Do AI visibility tools cover Google AI Overviews? Four of the six other products in the table name it on their own pages, and two say "Google AI" without specifying the surface. Saying it and measuring it are different claims, and this page records only the first: we have not tested anyone's coverage. Ours does not cover AI Overviews at all, and that is a real gap for buyers who need it. The surface has no official API, so if it matters to you, ask any vendor claiming it how the data is collected and what happens to your history when Google changes the format.
Why does a tool's number not match the one I get by asking ChatGPT myself? Because the answer changes between asks, so neither figure is the truth on its own. Our own three runs above produced 100%, 100% and 66.7% from an identical question set inside fifteen minutes. Compare ranges rather than points, and treat a mismatch as informative only when your range and the tool's number do not overlap at all.
Can I check my AI visibility for free before paying anyone? Yes, and you should. Several vendors in the table offer a free checker, ours included, and all of them are smoke tests rather than baselines: they confirm that your brand matching works and give you one number with a wide range. You can also do it with no tool at all by writing ten buyer questions that do not contain your brand name, asking each three times, and recording a yes or no. That produces a mention rate with a denominator you chose, which is more than most dashboards will show you.
Why is there no ranking or score on this page? Because we sell a competing product, and a ranking from a competitor is an advertisement with a table around it. The four columns are chosen so that you can rebuild the whole comparison yourself from the source URLs below, disagree with our reading of any cell, and reach a different shortlist than we would.
Sources
Every cell above traces to one of these pages, all read on 2026-09-20. No affiliate links, no paid placement, no ordering by anything other than the alphabet.
- Ahrefs Brand Radar:
ahrefs.com/brand-radar,ahrefs.com/ai-visibility-checker - Frase:
frase.io/features/ai-tracking,frase.io/pricing - Otterly.ai:
otterly.ai,otterly.ai/pricing,help.otterly.ai/search-prompt-monitoring - Peec AI:
peec.ai,peec.ai/pricing,peec.ai/ai-instructions,docs.peec.ai/setting-up-your-prompts - Profound:
tryprofound.com,tryprofound.com/pricing,help.tryprofound.com/articles/3443229936-answer-engine-insights-overview - RedClaw Labs: our own methodology and AI visibility tracking pages, plus the run records for the three Miro runs of 2026-09-20
- Semrush:
semrush.com/pricing/ai/,semrush.com/kb/1493-ai-visibility-toolkit
Prices and engine lists in this category change often. If you are reading this some months after the date at the top, treat the table as a list of questions to ask rather than as current fact, and open the source pages.