Are you in the answer, or is your competitor?
AI visibility tracking asks the questions your buyers actually type into AI assistants, several times a week on three engines, and records whether you were cited, merely named, or replaced by a rival. Then it names the page that lost and rewrites it.
The crawler-access check needs no account. Prompt tracking runs on your plan credits.
Example data. Every figure ships with its interval and its sample count.
Can AI engines reach you, and is what they find worth citing?
Those are two different problems with two different fixes, and most tools only look at the second. Reachability is checked from your robots.txt and your rendered HTML; citability is checked from the answers themselves.
1 · Can they reach you?
We read your robots.txt crawler by crawler and check whether the page renders without JavaScript. A page an AI engine cannot fetch cannot be cited, however good the writing is. It is also the cheapest problem on this list to fix.
2 · Which crawler is which?
Blocking a training crawler costs you nothing. Blocking a retrieval crawler removes you from that engine entirely. GPTBot is training; OAI-SearchBot is what ChatGPT actually cites. We name each one and what blocking it costs you.
3 · Is the content citable?
Your library ranked worst-first by GEO score, with the failing checks named — so you know what to fix rather than only that something is wrong. The AI-traffic tab shows which bots actually fetched you, read from your own server logs.
What does one weekly run actually return?
One row per prompt, one column per engine, and one of five outcomes in each cell. The table below compares the same four prompts across all three engines, with the competitors that appeared instead of you in the last column.
| Prompt | ChatGPT | Gemini | AI Overviews | Who else appears |
|---|---|---|---|---|
| best boiler service company in Leeds | cited | named | absent | Britherm, Checkatrade |
| how much does a heat pump install cost uk | absent | absent | rival only | Vaillant UK, Energy Saving Trust |
| acme hvac vs britherm | named | rival only | no answer | Britherm |
| emergency boiler repair west yorkshire | absent | named | absent | Trustpilot, Britherm |
We refuse to call a change that is not there
AI answers vary run to run, so we ask each prompt several times per platform over a 28-day rolling window and show a range, not a rank. When two ranges overlap, the interface says “no change we can detect” — even when the midpoint moved. That refusal is the feature: a tool that reports noise as progress is worse than no tool.
“No answer” is not a miss
Google does not render an AI Overview for every query. When it does not, we record no answer and exclude the prompt from the percentage. Counting it as a miss would show a visibility collapse that never happened, which is exactly the mistake that makes these dashboards untrusted.
What do you do with the answer once you have it?
Everyone else tells you that you are invisible. Every cited URL here is joined to a crawled inventory of your own pages, so a miss stops being a percentage and becomes a page with a named defect. The prompt’s intent then picks the matching GEO template and the generation engine rewrites that page. Next week’s run measures whether it worked — attribution no monitoring vendor can offer, because none of them owns a writer.
| Your page | Citations | Verdict |
|---|---|---|
| /services/boiler-servicing | 7 | winning |
| /heat-pumps | 0 | no FAQ schema · answer at 780 words |
| /pricing/landlord-certificates | 0 | table only, no prose to quote |
Half the time the winner is not a competitor at all
We aggregate every cited domain, not only the ones you thought to track. Customers routinely discover that a directory or an authority site is taking the answer — which means the work is earning a listing, not writing a page. That is a different plan, and you cannot make it without the data.
| britherm.co.uk | 19 | competitor |
|---|---|---|
| checkatrade.com | 11 | not tracked — directory |
| energysavingtrust.org.uk | 8 | not tracked — authority |
| acme-hvac.co.uk | 7 | you |
What does AI visibility tracking cost?
The crawler-access check is free and needs no account. Prompt tracking runs on the same credits as everything else you generate — every check calls a real answer engine, so it is metered rather than unlimited. One plan, one balance, no second bill.
Yearly plans are billed once for twelve months, shown here per month. Cancel renewal any time.
Free
Try it on your own site
- ModelDeepSeek
- GEO score — all 16 checks, each showing its rule
- Answer-first passage writing, built to be quoted
- Every content template, and the full editor
- Critic pass — a second look before you read it
- Images chosen and inserted for you, free and licensed
- Internal links and an FAQ block, added automatically
- Connect WordPress and publish for real
- One weekly schedule — watch it run itself
- Free AI-visibility check, unlimited, no signup
Pro
Set it running
- Models4 providers, 5 models
- Live research and source citations as you write
- GEO auto-optimise, and the chunk view that shows what an engine lifts
- Unlimited schedules — daily, weekly, drafts or live
- AI-crawler report: which bots fetch you, which pages they never touch
- Search Console AI impressions beside your own crawler data
- Custom templates in your brand voice
- Multiple projects and publish destinations
- Priority support
BYOK
Bring your own API key
- ModelsAny model, your key
- No credit ceiling — volume is between you and your provider
- No markup on a single token
- Unlimited schedules on every model
- Live research and citations
- Custom templates in your brand voice
- Multiple projects and publish destinations
A check is one prompt asked once on one engine. Asking 25 prompts on three engines three times a week is roughly 975 checks a month, so plan your prompt list against your credit balance rather than assuming it is free.
What people ask before they buy this
Is this the real ChatGPT?
It is the OpenAI API with web search enabled, and it is labelled that way everywhere in the product. There is no official API for the consumer product at chatgpt.com, and scraping it would breach OpenAI’s terms, so we do not. Any vendor claiming to read the real consumer product is overselling.
Why do you show a range instead of a rank?
Because AI answers vary by session, personalisation and location. A single number implies a precision that does not exist. We ask each prompt several times per platform across a 28-day rolling window and print the interval and the sample count beside every figure.
What happens when Google shows no AI Overview?
It is recorded as "no answer" and excluded from the visibility percentage rather than counted as a miss. Google not rendering an AI Overview for a prompt is a fact about Google’s coverage, not about your site, and conflating the two shows a collapse that never happened.
Do I need to pay anything to start?
No. The crawler-access check is free, unlimited and needs no account: it reads your robots.txt crawler by crawler and checks whether your pages render without JavaScript. Prompt tracking spends credits because every run calls three separate providers for real.
How many credits does tracking use?
One check is one prompt asked once on one engine, and it costs roughly what the provider charges us — a fraction of a cent, metered the same way research and generation already are. Twenty-five prompts on three engines three times a week is about 975 checks a month, which is why the prompt list is yours to choose rather than a fixed allowance.
Start with the free half.
Find out whether ChatGPT, Gemini, Claude and Perplexity can fetch your pages at all. No account, and it is the cheapest problem you will ever fix.