AI visibility, mended
See where you’re missing from AI answers — and get it fixed.
When your customers ask ChatGPT, Perplexity or Gemini for a recommendation, either you’re in the answer or a competitor is. Answermend shows you which — then gives you the files that fix it — ready to paste, or to hand to whoever built your site.
For e-commerce and product brands.
Example data. Northlight is an invented brand on northlight.example — no customer’s scan appears anywhere on this site. The screen is the shipping report: 8 prompts, 4 assistants, 1 run each — 32 runs attempted, 30 came back.
The dashed Not measured tile is the one to check a vendor on: a check that does not run is counted there, separately — excluded from every denominator and never shown as a zero or a miss. On this example scan 2 of its 32 runs did not come back, so it reads 2 — and those 2 sit in none of the three denominators beside it.
Checks the answers in
- ChatGPT
- Gemini
- Claude Pro
- Perplexity Pro
When AI answers your customers
Either you’re in the answer, or a competitor is.
Your customers stopped scrolling ten blue links. They ask, and they act on the answer. These two screens are how you find out which way that went — shown here with example figures.
Example data — the shipping share-of-voice screen on the invented brand Northlight. Every bar states its own denominator.
The rivals come out of the answers, not out of a list you typed.
These domains were read out of the very answers the scan collected — who was named, and cited, on the questions about you. And no bar is a bare percentage: named on 6 of 8 searched questions carries its denominator with it, and a question no engine answered is in nobody’s denominator.
Example data — the shipping sentiment screen on the invented brand Northlight. Every line states the runs it was read over.
One answer is an anecdote, not a sentiment.
This screen will not turn a single reply into a mood score. From fewer than two runs it publishes no distribution at all, and the label one run got is shown as what one run said, not as what the answers say. A run the provider never returned is not measured — never neutral, never a zero — so the silence never quietly improves your average.
Four things can happen when an engine answers a question about you.
AI visibility isn’t binary. Three of the four are findings. One is us admitting we don’t know.
whats the best crm for a small law firm
A page on your site was fetched and used as a source. This is the strong result.
The model named you from memory, without opening your site. Your brand is known; your site did nothing.
The answer ran, and never named you. Someone else got the recommendation.
The check didn’t return — a rate limit, a failure, a cap. Reported as unrun, with the reason, and left out of every count.
Inside the report
This is the screen. The figures on it are an example.
Northlight is an invented brand on northlight.example and every number below is made up. The layout, the four states and the arithmetic are the real ones.
Every cell states the runs it was counted over, and a run that did not come back is printed as not measured in its own cell, with a why beneath it, rather than folded into a count — never a zero, never a miss.
Example data throughout. No customer’s scan appears anywhere on this site — that record is confidential. On your own scan the questions are built from your homepage, and the row marked you is your domain.
The mend
Don’t just find the problem. Ship the fix.
Answermend turns a gap into a file you can paste — or hand to whoever built your site.
- The gapYou are not in the answer
- The fileA block of text you can paste
- In placeOr hand it to whoever built your site
Example data — one item from the shipping What-to-fix screen, on the invented brand Northlight. Every item names the questions it was derived from: the two listed at its foot are the ones this finding comes out of.
We ship the artifact and re-measure. We do not promise a mention — no vendor controls what a model answers.
Free tools. The result on screen.
Seven of the twenty are below: four of the eight generators that run in your tab, and three site checks that fetch your live site from our server.
Generators — in your browser Nothing is sent anywhere.
FAQ Schema Generator
Turn your FAQs into something AI assistants can quote accurately. Paste one block into your site — or hand it to whoever built it.
Open the tool → runs in your browserOrganization Schema
Make sure AI knows who you are: your real name, where you are, your official profiles. Stops it confusing you with a similarly-named business.
Open the tool → runs in your browserrobots.txt Generator
Decide which AI companies are allowed to read your site, and which are not. One small file, written correctly.
Open the tool → runs in your browserllms.txt Generator
A short summary of your site, written for AI to read. Worth an hour — though no AI company has confirmed they use it, and we won’t pretend otherwise.
Open the tool → runs in your browserSite checks — server-side These fetch your site from our server.
AI Crawler Checker
Whether you are accidentally blocking AI from reading your site. Plenty of businesses are, without knowing it.
Open the tool → runs on our serverRender Check
See your pages the way AI sees them. Sites that need JavaScript to show their content can look empty to it.
Open the tool → runs on our serverAI Readiness Score
One check covering all of the above, so you know what to fix first — and anything we could not test is left out of the score, never counted against you.
Open the tool → runs on our serverWhy trust the number
We’d rather show you a gap than a number we made up.
No score out of 100
Cited and mentioned are different facts. Adding them into one number would be inventing precision.
Unrun is never zero
A rate limit isn’t an absence. It’s reported with its reason and left out of every count.
One run says one run
One pass per question, labelled as one run rather than as an average. Repeated runs are what the paid plans add.
No guaranteed mention
No vendor controls what a model answers. Anyone promising a mention is selling you a coin flip.
What the report prints
What it never prints
The whole method is written down once — the questions we ask, the four states, and why an unrun check is never shown as a zero — on how it works.
Start free. Pay when it’s worth paying for.
Free tells you whether AI knows you. Starter watches one brand daily and hands you the fixes.
Free
$0
no card
For finding out whether AI assistants know you exist.
- Available today: Three scans in any 30 days, no card
- Available today: Up to 5 questions, written from your homepage
- Available today: Checked on ChatGPT and Gemini
- Available today: See who AI names instead of you
Creates your account. Three scans in any 30 days.
Starter
$49/mo
or $441 a year — three months free
Citations measured on both assistants every plan covers, a daily check, and the fix files. Pro adds breadth. Start here, and move up any time.
For one brand that wants watching every day, and fixing.
- Switched on, not yet proven: Checked every day: Gemini daily, ChatGPT weeklynot yet proven
- Available today: Citations measured on ChatGPT and Gemini
- Available today: Fix files: schema, llms.txt, robots.txt, FAQ blocks
- Available today: Competitors named, with the questions they win
Sign in, then pay on Paddle’s checkout from your account page.
Pro
$99/mo
or $891 a year
For a brand competing across many questions and four assistants.
- Available today: Everything in Starter, across 40 questions
- Switched on, not yet proven: Adds Claude (monthly) and Perplexity (every third day)not yet proven
- Available today: Drafted pages for the questions you’re missing
- Switched on, not yet proven: An email when an AI answer about you changesnot yet proven
Sign in, then pay on Paddle’s checkout from your account page.
Thirty days to change your mind. Write to us inside thirty days and the whole amount comes back — no conditions, and you do not have to say why. Cancel any time by one email or from your account page, and you keep access to the end of the period you paid for. Free needs no card.
Not yet proven, so tagged above: Starter’s schedule and the Pro alert email have never fired; Claude and Perplexity have not run twice yet. The daily run has fired every day since 6 Sep 2026 on our own Pro account. Everything else runs today, when you press the button.
Fair questions.
What’s the difference between mentioned and cited?
Mentioned means the model named your brand from memory. Cited means it fetched and linked your page as a source. The first is a reputation signal, the second is a content and crawling problem — they need different fixes.
Which engines do you check?
Every plan covers ChatGPT and Gemini; Pro adds Claude and Perplexity. Google AI Overviews has no public API and nothing measures it today — named here rather than quietly left off the list. Any engine we cannot measure is named as unmeasured rather than guessed at.
Can you guarantee my brand gets mentioned?
No — and nobody can. What we can do is show you exactly where you’re missing, and fix what those systems actually read on your site.
Check where you stand, then fix it.
Every tool on the shelf is open.