Skip to content
EchoVerseTen focused tools

Measurement

How to check if ChatGPT cites your website

Short answer

There is no Search Console for AI citations, so you have to sample. Write a fixed list of 20 prompts a real buyer would type, run them in ChatGPT on a schedule, and record for each whether you were cited, mentioned without a link, or absent. The trend across repeated runs is the signal; any single run is noise.

By The EchoVerse teamUpdated 4 min read

Why a single check tells you nothing

Answer engines are non-deterministic. The same prompt can return different sources on different runs, in different sessions, and for different accounts. Checking once and concluding you are invisible — or that you have won — is reading noise.

What is measurable is the trend across a fixed prompt list, run the same way on a schedule.

The method

1. Write 20 prompts a buyer would actually type. Problems and comparisons, not your brand name. "Best invoicing app for contractors" is a real prompt. "Is BillingDone good" is a prompt only someone who already knows you would type, and it measures nothing about discovery.

2. Record three outcomes, not one. For each prompt:

  • Cited — you appear as a linked source
  • Mentioned — you are named in the text with no link
  • Absent — no appearance

Those are genuinely different results. Collapsing them into one number is how tools inflate their metrics.

3. Note who was cited instead. This is the most useful column and the one people skip. The domains that appear are your actual competitive set, and they are frequently not who you assumed.

4. Repeat monthly, identically. Same prompts, same phrasing. Changing the list breaks the comparison.

Cross-check with your analytics

Referral traffic from AI surfaces appears in analytics like any other referrer. Segment it rather than pooling it: visitors arriving from an assistant have usually had their question answered already, so they arrive further along and behave differently from search traffic. Judging them by search benchmarks will make good traffic look bad.

Volumes are typically small. Do not read a low number as failure on its own.

What to do with a bad result

If you are consistently absent, work in this order:

  1. Confirm your content is in the served HTML. View source, not devtools. If the body

    2. copy only appears after JavaScript runs, most retrievers see an empty shell. This is the single most common cause and the least discussed.

  2. Confirm the crawlers are not blocked — in robots.txt and at the WAF layer.
  3. Read your opening paragraph cold. Does it answer the question in the heading,

    4. within about 80 words, without pronouns pointing at earlier text?

  4. Open the pages that were cited instead. The pattern is usually obvious within

    5. minutes, and it is usually structural rather than a matter of authority.

When to buy a tool

When the manual version becomes the bottleneck — more prompts than you will run by hand, or several markets to track. A spreadsheet on a monthly cadence beats a dashboard you stop opening, and it costs nothing to start.

Common questions

Can I see AI citations in Google Search Console?
No. Search Console covers Google Search. It does not report ChatGPT, Perplexity or Claude citations, and there is currently no equivalent console for them, which is why sampling a fixed prompt list is the practical method.
Why do I get different sources each time I run the same prompt?
Answer engines are non-deterministic, and results vary by session, account and time. That is exactly why a single check is unreliable and why you track a trend across a fixed prompt list instead.
Does referral traffic from ChatGPT show in analytics?
Yes, as a referrer like any other. Segment it separately, since visitors arriving from an assistant typically have their question already answered and behave differently from search visitors.
All articles