How to check if your website appears in ChatGPT
Last updated:
Short answer: ask the assistants the questions your customers would ask, without naming your business, and see whether you come up. If you do not, the cause is usually one of three things - the assistant cannot reach your pages, cannot extract anything useful from them, or nothing on the wider web says you exist. Each has a different fix, and only one of them is quick.
The test that actually tells you something
Most people test this wrong. They open ChatGPT, type “what is Acme Roofing?”, get a decent answer, and conclude they are visible. That question proves almost nothing: you handed the model the answer inside the question.
The two questions are completely different:
- Knowledge questionsname you - “What is Acme Roofing?” A good answer means the assistant can find or recall something about you. Useful, but nobody discovers a business this way.
- Discovery questionsdo not - “Who are the best roofers in Leeds?” This is how customers actually search, and it is the only one that measures whether you are being recommended.
A business can score perfectly on the first and zero on the second. That gap is the real measurement, and it is the one worth tracking.
How to run the test yourself
- Write five questions a customer would actually type, with no brand names in them. Include your category and your location.
- Ask each one in ChatGPT, Claude, Gemini and Perplexity separately. They use different indexes and will not agree.
- Use a fresh chat each time, and turn off personalisation or memory if you can. Otherwise you are measuring your own history, not reality.
- Record whether you appeared, and whether the assistant linked you. A citation is worth far more than a passing mention.
- Repeat monthly. A single result is noise; the trend is the signal.
Our free checkup automates the unprompted half of this: it generates discovery questions from your own page content, asks live AI engines, and reports how often you came up without being named. It is the “came up unprompted” figure that matters.
Why you are probably not showing up
There are three main routes into an AI answer, and they behave very differently. Knowing which one is failing tells you what to do.
1. The assistant cannot reach you
The fastest thing to rule out, and more common than people expect. If your robots.txt blocks OAI-SearchBot, Claude-SearchBot or PerplexityBot, you have removed yourself from those answers - often without realising, because plenty of “block AI” snippets circulating online do exactly that. Our AI crawler guide covers the main bots and what blocking each one costs. Fix time: minutes.
2. The assistant can reach you but gets nothing useful
A crawler that is allowed in still leaves empty-handed if your content only appears after JavaScript runs, your headings are styled divs rather than real ones, your key facts live in images, or the page has 40 words and a hero video. Retrieval systems work with extractable text. Fix time: days to weeks. This is what our AI Search Readiness category measures - see the full check list.
3. Nothing on the web says you exist
This is the real one, and the reason quick fixes disappoint. Assistants answer discovery questions by drawing on what the web says about a category - roundups, forums, directories, reviews, comparisons. If no source outside your own website mentions you, there is nothing for the model to surface, no matter how clean your markup is. Your own site claiming you are the best is not evidence. Fix time: months.
What actually moves the needle
- Be reachable - allow the search and user-initiated crawlers. Non-negotiable, and free.
- Be extractable- server-rendered text, real headings, specific facts. Say “fixed-price roof repairs in Leeds from £180”, not “quality you can trust”.
- Answer real questions on your own pages. Pages built around the question a customer asks are the ones retrieval finds.
- Get mentioned elsewhere - industry roundups, directories, forums, genuine reviews. This is slow, unglamorous, and the highest-value thing on this list.
- Stay in conventional search. ChatGPT search, Perplexity and Gemini grounding all lean on web indexes. Ordinary SEO is the substrate, not a separate discipline.
Honest timelines
- Days: unblocking crawlers, fixing extractability. Live-retrieval answers can pick these up quickly.
- Weeks to months: new pages being indexed and starting to surface for discovery questions.
- A year or more:becoming part of what a model “knows” without looking anything up. That follows from the web referencing you over time, and cannot be bought or engineered directly.
Anyone offering to get you into ChatGPT's training data on a timeline is selling something. What is genuinely achievable, and worth doing, is being the page that gets retrieved and cited when someone asks.