Choose an AI visibility tool by checking whose questions it asks, how many answers sit behind each figure, whether it repeats each question, which assistants and countries it covers, whether you can open every full answer and source, how long it keeps history, and what one tracked question costs. Then test it for a week on your own questions.
Need the answers for your own brand? The AI visibility audit puts 20 to 40 of your buyers’ questions to four AI models and records who they name instead of you.
- Every tracker multiplies questions by assistants by runs. Ask how many answers sit behind each percentage before you trust it.
- Answers change between runs: in AIGROW research the first business named changed in 55.8% of questions, so one answer per question is a single draw.
- Check how answers are collected (app or API, web search on or off, which country) and keep those settings fixed month to month.
- Open the full answers. A tool that only shows a score cannot be checked, and an answer that names nobody must stay in the count.
- Work out the price per answer. Peec AI publishes it: 50 prompts on 3 models daily make 4,500 answers for €85 a month.
How to evaluate an AI visibility tool
An AI visibility tool, also called an AI visibility tracker, puts a set of buyer questions to assistants such as ChatGPT, Gemini, Perplexity and Google's AI Overviews on a schedule, then counts how often the answers name your business, link to your site or describe it correctly. To judge one, check seven things: whose questions it asks, how many answers sit behind each number, whether it asks the same question more than once, which assistants and countries it covers, whether you can open every full answer, how long it keeps history, and what one tracked question costs. Then test it for a week on your own questions before you pay.
The seven checks, with the question to put to each vendor:
| Check | Ask the vendor | What a good answer looks like |
|---|---|---|
| Questions | Can I write my own questions, in my customers' language? | Yes, with branded questions kept in a separate group |
| Sample | How many answers sit behind each percentage? | The count is printed next to every figure |
| Repeat runs | Is each question asked more than once per period? | Daily, or several runs, with the number shown |
| Coverage | Which assistants, countries and languages, and which cost extra? | The ones your buyers use, in their language |
| Collection | App or API, web search on or off, from which country? | Written down and kept the same every month |
| Evidence | Can I open the full answer and the pages it cited? | Every answer, dated, and exportable |
| Price unit | What does one tracked question cost a month? | A price per prompt or per answer you can work out |
This guide is about trackers that count assistant answers. For the wider set of AI SEO tools, from content to technical audits, see the AI SEO tools guide with prices.
What a tracker measures, and what it only estimates
Every tracker multiplies three numbers: questions (vendors call them prompts), assistants, and runs. Peec AI spells it out on its pricing page: 25 prompts across 3 models for 30 days make 2,250 AI answers. Those answers are the evidence.
Every chart, visibility score and share of voice is a count over them.
The same question gives different answers depending on how it is asked, so find out how the tool collects them:
- App or API. SE Ranking says its tracker captures responses the way a real user would see them in the interface. Other tools call the assistants through their APIs, which is cheaper and easier to repeat but skips what the app adds. AIGROW's free scan uses the APIs, without web search.
- Web search on or off. OpenAI's help page says ChatGPT searches the web when a question needs current information, and may rewrite the question into several queries sent to its search partners. An answer with search can name different businesses from one without it.
- Location and memory. The same page says ChatGPT uses an approximate location from your IP address and, with memory on, saved memories when it rewrites a search. A tracker running from one server sees one version of that.
Some free tools do not ask the assistants at all. SEOcrawl's free checker calls its result "an estimate, not a live measurement": a model judges your likely visibility from your homepage. That can be a quick first idea.
A count of real answers is a different thing, so before you compare two tools' numbers, ask which kind each one is.
Why one answer per question is not enough
Assistants rarely give the same list twice. In AIGROW's study of AI recommendations, run without web search in September 2026, the first business named changed in 55.8% of questions between runs (13,536 of 24,269), and 54.3% of the businesses named appeared in one answer only. In the October 2026 AIGROW Index, asked in each country's language, pages in Germany repeated 50% of their names on a rerun and pages in the Netherlands 54%.
So the number of answers decides how much a percentage means. How often the tool asks sets that number:
| How often each question is asked | Answers per question, per assistant, per month |
|---|---|
| Once a month | 1 |
| Once a week | 4 or 5 |
| Once a day | About 30 |
| Once a day, three runs | About 90 |
Worked example (invented numbers). A tracker says ChatGPT named you in 5 of 30 daily answers to one question: 17%. The 95% interval runs from 7% to 34%, so a move to 20% next month tells you nothing. Pool ten questions and the same 17% from 50 of 300 answers narrows to 13% to 21%.
The AI visibility metrics guide has the formulas and an interval table.
Check the denominator too. An answer that names no business is a real answer and belongs in the count; a tool that drops it reports a higher rate than you have. Asked for a local business without web search in the AIGROW Index, Claude named no one in 58% of its answers and Grok in 76%.
The AI Recommendation Rate is defined with those empty answers kept in.
Which assistants, countries and languages it covers
Most trackers include ChatGPT and at least one of Google's AI features in the base price; Claude is often missing or costs extra. What each official page listed on 20 October 2026:
| Tool | Assistants in the plan | Extra cost or higher plan |
|---|---|---|
| Ahrefs Brand Radar | AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot | Claude in custom prompts, using 8 checks per update |
| Otterly.AI | ChatGPT, AI Overviews, Perplexity, Microsoft Copilot | AI Mode, Gemini and Claude as monthly add-ons |
| Peec AI | 3 models you choose, from ChatGPT, AI Mode, AI Overviews, Copilot, Gemini, Naver AI | Extra models as paid add-ons, Perplexity among them; Claude and Grok on Enterprise |
| Profound | ChatGPT, Gemini and AI Overviews on the trial | Up to 9 answer engines on Enterprise |
| Semrush AI Visibility Toolkit | Platforms such as Google AI Mode and ChatGPT | 50 more prompts for $60 a month |
| SE Ranking | AI Overviews, AI Mode, ChatGPT, Gemini | Plans list 5 LLMs to track; custom limits on Enterprise |
| Sistrix | AI Overviews, AI Mode, ChatGPT, Perplexity | Included while in beta |
Pay for the assistants your buyers use. A plumber's customers ask ChatGPT and Google; a software buyer may also ask Claude or Perplexity.
Language and country matter as much. A customer in Utrecht asks for a good accountant for a sole trader in Dutch, and gets a Dutch answer with Dutch firms. Ask each vendor whether it sets the country and language of the session or only writes the city into the question, and whether a market costs extra.
Profound's trial covers 1 language and 1 region; Peec AI lists multi-country tracking on its Advanced plan.
Full answers, sources and history
A percentage you cannot open cannot be checked. Before you buy, ask to see four things:
- The full answer behind every count, with the date, the assistant and the settings. Open ten and recount them: did the tool mark you as named where you really appear, and did it catch a misspelt version of your name?
- The pages each answer cited, when web search was on. Those are the sites to work on next. AIGROW's AI source statistics show which sites assistants read for each trade.
- History per question. Profound's trial keeps no history and its Enterprise plan keeps all of it. If a tool keeps only the latest answer, you cannot see when a change began.
- Exports. Semrush's toolkit exports up to 1,000 rows per CSV, 10 exports a day. Profound lists CSV and JSON exports and an API on Enterprise. If you cannot get your answers out, your history stays with the vendor.
Accuracy deserves its own line. Being named with the wrong prices or opening hours can cost more than not being named. Peec AI lists fact-checking of prompts against facts you set, and AIGROW's paid monitors check assistant claims against your saved business facts.
Both are comparisons against what you supply, so keep those facts up to date.
What AI visibility tools cost in October 2026
Prices as each tool's own page showed them on 20 October 2026, monthly billing, in the currency the page showed: euros for Ahrefs, Otterly.AI, Peec AI, SE Ranking and Sistrix, dollars for Profound and Semrush. Some pages switch currency by location. They change often, so check the page before you buy.
| Tool | Published entry price | What the price buys | Trial, free or included option |
|---|---|---|---|
| Ahrefs Brand Radar | Custom prompts from €47 a month; AI Visibility Index from €179 | Checks: 1 per prompt, platform, location and update; €654 a month for all models | Daily custom prompts in paid Ahrefs plans (Lite 5, Standard 10, Advanced 20) |
| Otterly.AI | €29 (Lite, 15 prompts) | €189 for 100 prompts, €489 for 400, 15% off yearly | Free trial for new users, length not stated |
| Peec AI | €85 (Starter, 50 prompts) | 3 models daily, 4,500 answers a month; €205 for 150 prompts, €425 for 350 | Free trial offered, terms not stated |
| Profound | No published price for brands | Enterprise by quote; for agencies, client workspaces at $399 a month each, 100 prompts daily | 7 days: 50 fixed prompts a day, 3 engines, no history or exports |
| Semrush AI Visibility Toolkit | $99 | 25 prompts tracked daily, 1 project | No free trial |
| SE Ranking | €109 (Core) | 100 prompts tracked daily inside an SEO plan; €235 for 250 | 14 days, 20 AI prompts a day |
| Sistrix | €119 (Start) | AI analysis free in beta: 100 prompts a month, up to 1,500 on Premium | 14 days free |
Profound's trial runs a recommended prompt set you cannot change, so it cannot tell you how often your business is named for your own questions.
Price per answer. Divide the monthly price by prompts × assistants × runs per month. Peec AI's Starter plan gives the clearest case because the page prints the answer count: €85 for 4,500 answers is just under 2 cents an answer. An invented $100 plan with 25 prompts on 2 assistants, asked daily, gives 1,500 answers, about 7 cents each.
Ask any vendor that does not publish the count for the three numbers, then do the sum.
For a wider price table that also covers BrightLocal, Scrunch, AthenaHQ and Surfer, see the AI SEO tools guide, and AIGROW's comparison of AI visibility tools.
Want to know where your business stands? The free checker shows whether ChatGPT, Claude, Gemini and Grok name it.
Free ways to check AI visibility, with or without an account
You can get a first reading without paying. What each free option gives:
| Option | Account | What you get |
|---|---|---|
| By hand | None | Ask 10 buyer questions in each assistant's free app, signed out, and save the answers in a sheet |
| Google Search Console | Google account | Clicks and impressions from AI Overviews and AI Mode, counted inside the Performance report |
| SEOcrawl free checker | None | Up to 3 reports a day; the page calls the result an estimate |
| SE Ranking free check | Not stated on the page | Up to 5 competitors, 5 attempts a day |
| AIGROW free AI scan | None | Five questions to the models behind four assistants, each rate with its 95% interval |
| AIGROW AI visibility checker | None | Two buyer questions to four assistants, every answer word for word |
Google's own page on AI features says sites shown in AI Overviews and AI Mode are counted in the Search Console Performance report under the Web search type. The metrics guide shows how to read that report and Bing's AI Performance report.
The free AI scan needs no account and no card. It reads your homepage, puts five questions to the models behind ChatGPT, Claude, Gemini and Grok through their APIs, without web search, and shows how often each names you. Under current limits it asks each question once per assistant, up to three times when limits allow, and the result says how many.
The free AI visibility checker takes about 30 seconds and needs no sign-up. It asks two buyer questions about your trade and city to ChatGPT, Claude, Gemini and Grok, once each, with no web search, and prints every answer word for word: whether you were named, in what place, and who was named instead. Two questions answered once are a snapshot, useful to read the actual answers before you choose a paid tracker.
A one-week test before you pay
Use a trial or a first month to test the tool on your own business:
- Write 20 questions your buyers ask, in their language, with no brand name in them. Keep 5 that name you or a competitor in a separate group.
- Ask five of them by hand in ChatGPT and Gemini and save the answers with the date.
- Load all 20 into the tool with the language and country your buyers use.
- After seven days, pick ten answers at random, read each in full and check the tool's marks.
- Count the answers that named no business and confirm they are in the denominator.
- Compare the tool's answers with yours from step 2. Different wording is normal. A completely different set of businesses means different settings, so ask which.
- Export everything. If the export fails or leaves out the answer text, note it.
- Work out the price per answer for the plan you would need.
Worked example (invented numbers). 20 questions × 3 assistants × 7 days = 420 answers. You are named in 63: 15%, with a 95% interval of 12% to 19%. ChatGPT alone named you in 21 of 140: 15%, interval 10% to 22%.
In step 4 you find 2 of 10 answers marked wrongly. That error is larger than the interval, so ask the vendor how it matches names before you trust any trend.
If you would rather have someone collect and check the answers, the AI visibility audit asks 20 to 40 of your buyer questions across four assistants, records every answer and the competitors named in your place, and ends with an ordered list of fixes.
Where AIGROW fits
AIGROW is one of these options, built for small businesses and the agencies that serve them. What its own pages describe:
- Free: the AI scan and the AI visibility checker above, both without an account.
- Monitoring plans: Watch checks 10 questions monthly; Act checks 30 questions weekly with three competitors. The paid monitor saves every answer and who it named instead of you. It checks assistant claims against your saved business facts, offers question suggestions from your own Search Console data, and tracks up to 2 markets on Watch or up to 6 on the higher plans, each in its own language. Operate and Agency add visits from AI assistants, from your GA4 export. Details on the plans page.
- Audit: the AI visibility audit, described above.
- Agencies: a white-label monthly report with your logo and a client link, on the agency plan.
Limits to know: the place is written into the questions, so the monitor does not reproduce a visitor's device location. Monitoring records model API answers; it cannot guarantee a ranking, recommendation or sale.
AIGROW measures itself the same way. In October 2026, across 358 answers to 30 questions in 11 languages, ChatGPT, Claude, Gemini and Grok named AIGROW in 4 (1.1%). Semrush, Profound and Otterly.AI were named most often instead. The study lists every question.
Sources
Official pages read on 20 October 2026:
- Ahrefs: Brand Radar
- Otterly.AI: Pricing
- Peec AI: Pricing
- Profound: Pricing
- Semrush Knowledge Base: AI Visibility Toolkit
- SE Ranking: AI Visibility Tracker
- SE Ranking: Subscription plans
- Sistrix: AI and chatbot analysis
- Sistrix: Pricing
- SEOcrawl: AI Visibility Checker
- OpenAI Help Center: Searching the web with ChatGPT
- Google Search Central: AI features and your website
AIGROW data:
- AI recommendations study, 2026
- AI Visibility Report, October 2026 edition
- Does AI recommend AIGROW?
- AI Recommendation Rate definition
The worked examples use invented numbers. Every interval in them is a 95% Wilson interval.
Check your own business in AI answers
Four assistants, your own buying prompts, five working days. If it turns out you are already named, the audit says so and costs you nothing further.
One fixed price, paid once
Frequently Asked Questions
