The short answer
What does an AI visibility tool do?
It asks AI engines the questions your buyers ask and records whether your brand is in the answer.
An AI visibility tool works like a rank tracker for AI answers. Instead of asking where you rank on Google, it asks whether ChatGPT, Claude, Google AI Overviews and Gemini name you when someone asks about your category, which competitors they name instead, and which pages they read to build the answer.
mentionedin.ai does that in seven steps. It asks each prompt 3 times per engine in one test, so the label comes from 3 answers, then turns what it finds into work on your own pages and tests again after you ship.
Step 1 of 7
How does brand setup work?
The setup reads your site and drafts a profile. You correct it.
The setup reads your home page and up to four key pages, pricing first, and drafts a brand profile: your name and the other ways people write it, your category, your audience, your market and 8 to 12 facts about you. A fact with a number in it is kept only when every number appears on your site.
You edit that into a truth file of 5 to 15 facts, the things an AI answer should get right about you, and confirm at least one competitor. Competitors are suggested from comparison pages on your own site and from search data, and the top five are switched on. The first test also shows which other brands the engines name instead of you, so you can add them.
Step 2 of 7
Which prompts should you track?
15 tracked prompts across the buying journey, plus two controls.
Prompts are generated from templates for each stage of buying: branded, pricing, comparison and discovery questions. You add the 3 to 5 questions you most want to win. Search volume is shown where it is available.
Prompts 5
named in a run not named
| Buyer prompt | ChatGPT | Claude | AI Overviews | Gemini |
|---|---|---|---|---|
| best client reporting tool for agenciesDiscovery3/12 runs | 1/3Weak | 2/3Moderate | 0/3Invisible | 0/3Invisible |
| AgencyAnalytics alternativeComparison1/12 runs | 0/3Invisible | 1/3Weak | 0/3Invisible | 0/3Invisible |
| white label reporting dashboardDiscovery9/12 runs | 3/3Strong | 3/3Strong | 1/3Weak | 2/3Moderate |
| how much does yourbrand costPricing9/12 runs | 3/3Strong | 3/3Strong | 0/3Invisible | 3/3Strong |
| best invoicing app for freelancers Control, no fixes0/12 runs | 0/3Invisible | 0/3Invisible | 0/3Invisible | 0/3Invisible |
best client reporting tool for agencies
Discovery3/12 runs
1/32/30/30/3AgencyAnalytics alternative
Comparison1/12 runs
0/31/30/30/3white label reporting dashboard
Discovery9/12 runs
3/33/31/32/3how much does yourbrand cost
Pricing9/12 runs
3/33/30/33/3best invoicing app for freelancers
Control, no fixes0/12 runs
0/30/30/30/3
Named in 22 of 48 runs on these prompts, least often on Google AI Overviews. The control prompt gets no fixes: when it moves, the engine moved, not you.
Each prompt shows whether engines usually search the web to answer it, and you can override that. It matters: an answer built from the engine's memory will not move when you change your site, at least not until a new model version is trained, so the prompts worth fixing for are the ones where the engine goes looking.
Two control prompts that no fix will ever target measure how much the engines drift on their own. They are left out of every headline number and do not count toward your tracked prompts. Prompt sets are versioned: an edit makes a new version, and a prompt with the same wording keeps its history.
Step 3 of 7
Why ask every engine each prompt 3 times?
One AI answer is a sample. The score comes from a consistency test.
Ask an engine the same question twice and you can get two different lists of brands. So a consistency test asks each prompt 3 times on each of the four engines, ChatGPT, Claude, Google AI Overviews and Gemini, which is 12 engine runs for one fully tested prompt. ChatGPT, Claude and Gemini are asked through their official APIs with web search on. Google AI Overviews is read from a live Google results page for the country set on your brand.
best client reporting tool for a small agency?
Run 1Named, 2nd
AgencyAnalytics, Your brand, Whatagraph, DashThis
Run 2Not named
AgencyAnalytics, Whatagraph, DashThis, Swydo
Run 3Named, 3rd
Whatagraph, AgencyAnalytics, Your brand
Why 3 and not more: in our calibration, 3 runs reached the same verdict as a test with more than three times the runs in 225 of 240 prompt and engine cells.
| Label | Named in | What it means |
|---|---|---|
| Strong | 3 of 3 runs | The engine names you every time. |
| Moderate | 2 of 3 runs | You are in most answers, not all. |
| Weak | 1 of 3 runs | You show up, but it could be luck. |
| Invisible | 0 of 3 runs | Every run completed, the engine did not answer purely from memory, and you were never named. |
Two more labels keep the record honest. Inconclusive means some runs failed or the engine never searched, so a zero proves nothing. Error means every run failed. Neither is shown as Invisible.
Your mention rate per engine, your share of voice against each competitor, the cited domains ranking and the accuracy check all come from these tests. Every run stores the full answer, the sources it cited, whether the engine searched and the engine version, so any number traces back to the answers behind it.
What counts
What counts as a mention?
Your name in an answer is not enough on its own.
- The run finished without an error.
- Your brand, or one of the ways people write it, is named in the answer.
- The answer is not just repeating your name back from the question.
- It is really you: confirmed by your domain, a fact from your truth file, or a check of the answer.
When an engine uses your name for a different company, it is not counted. It is shown apart as a mix up, so you can see where engines confuse you with someone else. Competitors are scored by the same rules on the same prompts.
Answers to your branded and pricing prompts are also checked against your truth file and shown as how many checked answers matched your facts. Sentiment is recorded too.
Between tests
What happens between consistency tests?
A weekly single run per prompt, shown as one sample and never as a score.
You can switch on a weekly spot check on the Checks page. It asks every prompt once on every engine and shows the result as a spot check with one sample, never as a rate or a label. Spot checks do not use your monthly consistency runs.
When a prompt that last tested Strong or Moderate misses two spot checks in a row, a 3 run consistency test on that engine is queued, at most twice per prompt per month. That test, not the spot check, decides whether the drop is real. Those confirmation tests do count toward your monthly consistency runs.
Step 4 of 7
Which pages do AI answers cite?
The sources behind every answer, ranked by how often engines cite them.
Every answer's cited sources are stored, so you get a ranked list of the domains AI answers read in your category, split into your own domains and third party ones, per engine. For Google AI Overviews, a cited domain is one on the results page: the overview's sources plus the top 10 organic results.
Most cited domains
Our own brand, day 0 test, 30 Sept 2026
- mentioned.aiCompetitora competitor's own site51 answers
- g2.comreview site44 answers
- mentionable.ai39 answers
- alternativeto.netalternatives list38 answers
Listicles that name competitors, not you
Source gaps, same test
seranking.com/blog/profound-alternatives
ProfoundPeec AIOtterlyYou: not namedCited byClaudescrunch.com/blog/best-answer-engine-optimization-aeo-generative-engine-optimization-geo-tools-2026
ProfoundPeec AIOtterlyScrunch AIYou: not namedCited byClaudeomnibound.ai/blog/best-aeo-tools-for-answer-engine-optimization
ProfoundPeec AIScrunch AIYou: not namedCited byClaude
Source gaps are the cited third party pages that name your competitors but not you. Each comes with its evidence, the prompts and answers that cited it, ranked by how often engines cite it. mentionedin.ai never contacts those sites or sends pitches on your behalf. The list tells your team where to work.
Step 5 of 7
How do you fix your own pages for AI answers?
The checks that decide whether an engine can read and quote you.
The page grader checks any URL on your own domain:
Before yourbrand.com/
Reporting, reimagined. The smarter way to wow your clients. Start your journey today.
First 50 Words: 0 of 4
Rewrite to copy First 50 Words
Yourbrand is client reporting software for small marketing agencies in the US and UK. It builds white label reports your clients open on their phone, from $49 a month.
| Check | Before 0/4 | After 4/4 |
|---|---|---|
| Who you are | ×fails | passes |
| What you do | ×fails | passes |
| Where you operate | ×fails | passes |
| What it costs | ×fails | passes |
| Readable without JavaScript (separate check) | passes | passes |
- First 50 Words: does the opening say who you are, what you do, where you work and what it costs?
- JSON-LD: is structured data present and valid in the HTML a crawler receives?
- Robots rules: does robots.txt allow the AI crawlers (GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, ChatGPT-User) and Google's Google-Extended control? PerplexityBot is checked so you stay crawlable; Perplexity answers are not tracked.
- llms.txt: does the site publish one? (Optional. Google has said it does not use the file, so it is a low priority fix.)
- Server rendering: can a crawler that does not run JavaScript read the copy?
- Category words: do the title and meta description name what you sell?
Any page that fails the First 50 Words check can get a rewritten opening in one click. The rewrite is built only from your truth file, every number in it must come from your truth file, current opening or profile, and it has to pass the same check. You see it before and after, with a copy button. You decide what ships; nothing is published to your site for you.
Step 6 of 7
What goes into the playbook?
Every finding becomes a job, grouped by week.
The playbook turns findings into checklist items: pages to create, pages to fix, First 50 Words rewrites, schema and robots fixes, llms.txt (optional, a low priority fix since Google has said it does not use the file), source gaps, listings and reviews. A prompt where engines rarely or never name you becomes a page to create in Week 1, and a prompt where you show up only some of the time becomes a page to strengthen in Month 1. A page that fails the grader becomes a page to fix.
Items are grouped into Week 1 for critical work, Week 2 for high and Month 1 for medium. A finding that stops showing up is marked no longer observed, never deleted, so the history stays.
Step 7 of 7
How do you prove a fix worked?
A dated entry for every change, then a new consistency test.
Every fix goes into a dated log with the page, what changed and an optional commit. Marking a playbook item done writes the entry for you, and you can add or edit entries by hand. The next consistency test shows whether the answer moved, and the report puts the fixes you shipped beside the results so you can compare them.
Prompt best client reporting tool for agencies
- Test 1, before the fix1/12ChatGPTClaudeAI OverviewsGemini
Fix loggedHomepage opening rewritten
Test 2, after the fix+8 runs9/12ChatGPTClaudeAI OverviewsGemini
Reports share at a private link that opens without an account, shows how often it was viewed and can be turned off at any time. A print layout saves it as a PDF.
Tested in public
Do we use it on our own brands?
The same tests, published as case studies anyone can read.
We track four of our own brands with mentionedin.ai, mentionedin.ai included, and publish the results for each brand once its first test has run. Each published case study shows the date it was measured and how many runs sit behind its numbers. In its one measured test, dated on its case study, AI answers did not name mentionedin.ai at all.
The published case study numbers were measured once, before consistency tests moved to 3 runs per prompt, and no fixes are logged on them yet. They are a starting point, not a before and after.
What it costs
What does mentionedin.ai cost?
$99 a month, or $79 a month billed yearly, with a 14 day trial.
Founder plan
$99a month
or $79 a month billed yearly ($948 a year). Prices exclude VAT.
Start free trial14 days free. A card is required at checkout; nothing is charged before the trial ends. Cancel anytime.
- 1 brand with its competitors
- 15 tracked prompts plus 2 control prompts
- 4 engines: ChatGPT, Claude, Google AI Overviews and Gemini
- A weekly spot check of every prompt on every engine, switched on in one click
- 204 consistency engine runs a month: enough to test all 17 once on every engine
- Page grader and First 50 Words rewrites for your own pages
- Playbook of pages to create, pages to fix and source gaps
- Intervention log and a shareable web report
Consistency pack
$19per pack, only when you need it
$19 for 99 more consistency engine runs, about 8 more prompts tested 3 times on every engine. Buy one for a month you want to test more, such as right after a batch of fixes. Packs are the only add on.
One prompt, fully tested
The plan's 204 runs test all 17 prompts, tracked plus controls, once a month this way. A pack adds about 8 more.
One engine run is one prompt asked once on one engine. The Founder plan includes 1 brand, 15 tracked prompts plus 2 control prompts, and 204 consistency engine runs a month: enough to test all 17 once on every engine. A pack of 99 more runs, about 8 prompts, costs $19 once and counts for the month you buy it.
The 14 day trial needs a card at checkout and nothing is charged before it ends.
Limits
What does mentionedin.ai not do?
Stated plainly, so you know what you are buying.
- It tracks four engines. Perplexity and other assistants are not tracked.
- It does not turn a single answer into a score.
- It does not publish to your website or CMS. You copy the rewrite and ship it.
- It does not write full pages or articles.
- It does not contact third party sites or send pitches.
- It does not send alerts or digest emails.
Questions people ask
How does an AI visibility tool work?
It asks AI engines the questions your buyers ask and records whether your brand is named, who is named instead and which pages the answers cite. mentionedin.ai asks ChatGPT, Claude, Google AI Overviews and Gemini each prompt 3 times per consistency test, labels the result, grades your own pages and tests again after you ship a fix.
Why does mentionedin.ai ask each engine 3 times?
AI answers change from one run to the next, so one answer is a sample, not a score. 3 runs per engine matched a much longer calibration test in 225 of 240 cells.
What do Strong, Moderate, Weak and Invisible mean?
Strong means named in 3 of 3 runs, Moderate 2 of 3, Weak 1 of 3, and Invisible 0 of 3 with every run completed and the engine searching at least once. Inconclusive and Error cover runs that failed or never searched.
Which AI engines does mentionedin.ai check?
Four: ChatGPT, Claude, Google AI Overviews and Gemini. Perplexity is not tracked.
Do I need to change my website to start?
No. Measuring needs only your domain. The page grader and the rewrites tell you what to change, and you decide what ships.
Does mentionedin.ai contact the sites that cite my competitors?
No. It lists them as source gaps with the prompts and answers that cited them, ranked by how often engines cite them. Nothing is sent to those sites on your behalf.
How many prompts does the $99 plan test?
15 tracked prompts plus 2 control prompts, and 204 consistency engine runs a month: enough to test all 17 once on every engine. A $19 pack adds 99 runs for that month.
Can I try mentionedin.ai first?
Yes. There is a 14 day trial with a card at checkout and nothing charged before it ends. The free check on the home page, without an account, asks 4 buyer questions 3 times each on ChatGPT, Claude and Google AI Overviews, shows one question free and opens the full report for a work email.