Is your brand cited by ChatGPT? How to check — and how to change the answer

When a prospect asks ChatGPT "best tools for X" and your competitor appears while you don't, you lost a deal you never knew existed. AI answers are the new first page — and unlike rankings, nobody sends you a report about them. Here's a monitoring loop you can run monthly without buying another tool.

Step 1 — Ask the machines what they say about you

Build a prompt list from your actual buyer questions — not vanity prompts:

• "best [your category] for [your ICP]" — the money question
• "[your brand] review" / "is [your brand] worth it"
• "[your brand] vs [competitor]"
• "how do I [the problem you solve]" — do you appear unprompted?

Run each in ChatGPT (with search enabled), Perplexity, and Gemini. Record three things per prompt: are you mentioned, are you cited with a link, and is what's said accurate. A simple spreadsheet with a monthly tab beats any dashboard for spotting drift — mentions appearing, disappearing, or mutating.

Two cautions: answers vary between runs, so treat one absence as noise and repeated absence as signal. And don't ask from an account with your brand in its history — memory contaminates the test.

Step 2 — Read the machine signals on your own site

The interviews tell you the output. Your traffic tells you the input — whether the machines are even reading the pages that would make them cite you:

• Is OAI-SearchBot crawling you? No index presence, no ChatGPT search citations — it's that mechanical.
• Are assistant fetchers (ChatGPT-User, Perplexity-User) hitting your pricing and docs? That's live conversations routing to you — the precursor of citations.
• Are AI referral clicks arriving from chatgpt.com or perplexity.ai? That's the citation working.

Crawled → fetched → cited → clicked. Wherever the chain breaks first is your actual problem, and it's usually earlier in the chain than people assume.

Step 3 — Fix what the AI reads, not what it says

You can't edit ChatGPT's opinion directly. You can edit its sources:

Your own pages carry surprising weight for brand questions — assistants routinely fetch the brand's site mid-answer. Make the About, pricing, and "what is [brand]" pages state facts plainly, in crawlable HTML, with current dates. How AI search decides who to cite covers the mechanics.

Third-party corroboration decides competitive prompts. "Best X" answers lean on comparison articles, directories, review sites, and forum threads. The entries you can legitimately influence — directory listings, honest comparison pages, an accurate Wikipedia-adjacent presence — matter more than another self-published superlative.

Inaccuracies get fixed at the source. When an answer states something wrong about you, ask which source it used (assistants usually tell you). Fix the source, or publish a clearly dated correction page — the machines re-read.

The uncomfortable part

Brand monitoring in AI answers punishes thin positioning fast: if your site never says plainly what you are, who you're for, and what you cost, the models either skip you or guess. Every fix in this loop is also just… good marketing hygiene. The AI layer didn't change what works; it removed the ability to fake it with ad spend.

Run the loop monthly, twenty minutes. The month you first show up in a "best tools" answer, you'll know exactly which change did it.