How to Check If Your Website Appears in ChatGPT, Perplexity and AI Overviews
A repeatable method for measuring AI mentions: which prompts to test, how to log results, and how to read the data honestly.
There is no analytics dashboard that tells you how often an AI assistant mentioned you. What exists is a manual method that takes twenty minutes a month and produces genuinely useful numbers. Here is the exact routine I use.
Step 1 — Build a fixed prompt set
Write ten to twenty prompts a real customer would type. Mix three types: brand queries (“is MKM.Fan legit”), category queries (“free AI visibility checker”) and problem queries (“how do I stop AI from ignoring my site”). Keep the list frozen so month-to-month comparisons mean something.
Step 2 — Run them in the same session type
Use logged-out, location-consistent sessions. Personalised chats, prior context and account history all bias answers. Run the same prompt in ChatGPT, Perplexity, Gemini and Google search with AI Overviews enabled.
Step 3 — Score four things per prompt
| Signal | How to record it | Why it matters |
|---|---|---|
| Mention | Was your brand named? yes/no | The headline number |
| Link | Was your URL cited? yes/no | Drives referral traffic and trust |
| Position | First, middle or last source | Earlier mentions are read more |
| Competitors | Who else appeared | Defines your real benchmark |
Step 4 — Compute two numbers
- Mention rate = prompts where you appeared ÷ total prompts.
- Share of voice = your mentions ÷ all brand mentions observed.
Two numbers, tracked monthly, are worth more than forty screenshots with no baseline.
Step 5 — Diagnose why you are missing
When you are absent, the cause is usually one of four things. Run the AI Visibility Checker to rule out the first two in under a minute:
- Crawler blocked in robots.txt, so the page was never retrievable.
- Page unreadable: no clear claim, thin content, JavaScript-only rendering.
- Entity ambiguity: the model cannot tell who you are or what you do.
- No corroboration: nobody else on the web says the same thing about you.
Ask the assistant directly: “which sources did you use for this answer?” On several engines this reveals the retrieval set, including domains you would never have guessed compete with you.
What good looks like over time
Month one is usually zero mentions, which is normal. Month two shows occasional appearances on brand queries. Month three to four is when category queries start including you, but only if you have added corroboration — guest contributions, open-source work, forum answers, documentation. The lag is real; do not abandon the routine at week three.
Do not buy “AI visibility” services that promise mentions inside model weights. No vendor can insert a brand into a trained model. What they can do is improve retrieval-side structure and publish citations elsewhere.
Frequently asked questions
Can I automate AI mention tracking?
Partially. You can script API queries, but each provider's answers vary by session and location, so a human-reviewed monthly sample stays more reliable.
Does appearing in ChatGPT drive traffic?
Some. ChatGPT increasingly links sources, and brand recall converts later through direct and branded search.
Why do answers change between two runs?
Generation is probabilistic and retrieval sets refresh. Track rates over many prompts rather than relying on single runs.