How to Know If ChatGPT or Perplexity Cites Your Website

To know whether ChatGPT or Perplexity cites your website, inspect a search-enabled answer, open its citations, and verify that a linked source resolves to your domain. Save the prompt and result. A brand name in the answer alone is a mention, not a citation.

I would start with a small, repeatable check before buying more tools or rewriting pages. Here is a workflow you can run manually and repeat after a meaningful update.

Separate mentions, citations, and visits

Measurement	What to record	What it does not prove
Brand mention	Your business is named in the answer	That your website was used as a source
Website citation	An answer cites a verified URL on your domain	That anyone clicked it
Referral visit	Your analytics attributes a visit to the platform	That you know which prompt or answer generated it

A third-party article about your business can support a brand mention while the citation belongs to that publisher. Record both accurately. If an answer links directly to your pricing page, save that exact page, not just the domain.

Choose five prompts to repeat

Use questions a potential customer might actually ask. Replace the brackets below with your category and target market. These are example prompts, not measured popular searches.

1. Category discovery: “Which [service or product category] providers should a [customer type] consider in [market]?”
2. Specific task: “How can a [customer type] solve [specific problem] in [market]?”
3. Buying criteria: “What should I compare before choosing a [service or product category]?”
4. Comparison: “Compare approaches to [task] for a business with [constraint].”
5. Brand check: “What does [your business] offer, and what sources support that information?”

Keep the branded prompt separate when reporting results. Naming your business in the question tests something different from appearing during category discovery. Do not add your URL to the discovery prompts, because that changes the test into a directed website lookup.

Run the check in ChatGPT and Perplexity

In ChatGPT, choose Search from the tools menu for a web-enabled answer. Open the inline citations and, where available, the Sources panel. OpenAI notes that this panel can contain cited sources and other relevant links, so record whether your URL is actually cited or only listed there. OpenAI’s search instructions explain the controls.

In Perplexity, submit the same question and inspect the numbered source citations. Open the source pages to confirm their URLs and what they support. This follows Perplexity’s description of its citations.

Use a fresh conversation for each prompt to reduce the effect of earlier messages. Note any personalization or location settings you know about. OpenAI documents that memory and location can influence search; a fresh conversation should not be described as a fully neutral test.

If a search fails or produces no inspectable answer, mark it as an incomplete check. Do not count a tool error as a confirmed absence.

Keep one row per completed answer

A spreadsheet is enough. Record these fields:

• Date and time, platform, and visible search mode or model.
• Exact prompt, requested market, and language.
• Brand mentioned: yes or no.
• Your domain cited: yes or no, plus the exact destination URL.
• Other cited domains and the sentence your citation supports.
• A saved answer, screenshot, or accessible conversation link.
• Any changed settings or uncertainty about the result.

For a first pass, five prompts on each of two platforms gives you ten planned checks. Report the actual completed count and keep each platform separate. If you later add prompts, flag the change so the next report does not imply the sample stayed identical.

What does a negative result mean?

“Not cited in these completed checks” is a defensible conclusion. “ChatGPT never cites us” is not. You have inspected a limited set of answers under particular conditions.

Repeat the same prompt set on a consistent schedule and after substantial page changes. Keep the earlier answers so you can compare actual evidence. Changing the questions each time makes it harder to tell whether your visibility changed or your test changed.

Our query fan-out guide explains why one broad question can involve several related information needs.

Turn the findings into one next action

If competitors are cited, read the specific pages and ask what useful evidence they provide. Check whether your equivalent page answers the question clearly and accurately. Improve a real gap rather than copying a competitor’s wording.

Review referral traffic separately in your analytics, with the reporting period and attribution limits visible. Exclude your own test clicks where your setup allows it. Citation counts and visits belong alongside each other, not combined into a single success claim.

To evaluate a product workflow, see DataWise AI Visibility. Keep the same evidence rules whether you check answers manually or use a tracker: source, prompt, platform, date, and a clear distinction between what was observed and what remains unknown.