How to check if ChatGPT mentions your brand

The shortest honest answer: open ChatGPT, ask your category question, and search the reply for your brand name. Then check whether your domain shows up in the answer's cited sources — that's a different question with a different answer. To do both repeatedly, across many phrasings, with history you can diff, fetch the answer as structured data instead of reading it in a browser.

The manual way (and why it stops working)

Open ChatGPT. Ask exactly what a prospective customer would ask — say, "best project management software for agencies" — and read what comes back. If Acme is named in the answer, Acme is mentioned. If the reply carries citations, expand them and look for acme.com. It takes a minute, it costs nothing, and for a one-off sanity check nothing beats it.

It stops working the second you need to do it twice.

• One query at a time. Genuine category coverage is twenty to fifty phrasings. You are not going to type all of them by hand every week, so you'll quietly shrink the check down to the two queries you happen to like.
• Answers vary between runs. Ask the same thing twice and you can get two different brand lists. A single reading can't tell you whether you gained ground or just drew a different sample.
• No history to diff. What you read today can't be compared with what it said in March, because March was never recorded. Every manual check starts from zero.
• Citations are easy to miss. The source that actually fed the answer is often behind a link you never expanded — so you conclude "we're invisible" when your own page was cited but not named.
• It doesn't fit in a report or a cron job. A browser reading can't go into a weekly deck, a client dashboard or a CI check without a human retyping it, which means it stops happening the week that human is busy.

Do it with one API call

AgentGEO fetches the answer for you and hands it back as JSON: the full answerText, plus sources[] with the title, URL and position of everything the answer cited. Same reply you'd have read in the browser, in a shape your code can test. Here's the raw call:

curl -s https://api.agentgeo.org/v1/fetches \
  -H "Authorization: Bearer ag_live_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"query": "best project management software for agencies", "surfaces": ["chatgpt"]}' \
  | jq '.answers[0] | {answerText, sources}'

Now the actual task. Test the two signals separately — a case-insensitive search of answerText for the brand name, and a scan of sources[] for your domain:

import requests

BRAND = "Acme"
DOMAIN = "acme.com"
QUERY = "best project management software for agencies"

resp = requests.post(
    "https://api.agentgeo.org/v1/fetches",
    json={"query": QUERY, "surfaces": ["chatgpt"]},
    headers={"Authorization": "Bearer ag_live_your_key_here"},
    timeout=200,  # live surfaces are slow - requests wait up to 180s
)
resp.raise_for_status()

for answer in resp.json()["answers"]:
    text = answer.get("answerText") or ""
    sources = answer.get("sources") or []

    # Signal 1 - is the brand named anywhere in the prose?
    mentioned = BRAND.lower() in text.lower()

    # Signal 2 - is our domain among the sources the answer cited?
    cited = [s for s in sources if DOMAIN in s["url"].lower()]

    print(answer["surfaceKey"])
    print(f"  mentioned in answer: {mentioned}")
    print(f"  cited as a source:   {bool(cited)}")
    for s in cited:
        print(f"    #{s['position']} {s['title']} - {s['url']}")

Two booleans, four possible outcomes — and each one means something different about what to go and fix:

Let your agent do it

If the check belongs in a conversation rather than a script — the way it usually does when you're mid-edit on the page in question — install the MCP server and ask in plain language:

claude mcp add agentgeo -- npx -y agentgeo-mcp --key ag_live_...

Then ask for the same two signals, in words:

Fetch what chatgpt answers for "best project management software for agencies". Tell me whether Acme appears in the answer text, and separately whether acme.com appears in the sources. If it's cited, give me the position and the page title.

The advantage isn't saving five lines of Python — it's that the agent already has your repo open. It can go from "you're mentioned but not cited" to "here's the paragraph on your pricing page that should have earned that citation" without you assembling the context by hand. The same server works in Claude Code, Cursor, Zed and any other MCP client.

Common problems

• Phrasing changes the answer. "Best project management software for agencies" and "project management tools for creative teams" are different queries with different brand lists. Test the phrasings your buyers actually use — pull them from your search console and your sales calls, not from your own vocabulary.
• One fetch is a sample, not a verdict. The same query run twice can return different brands. Before you report a change, re-run it; before you report a trend, run it on a schedule and compare records with the same query string.
• Short brand names collide. A substring test for "Acme" also matches "Acme Insurance" and "Acmegraph". Use a longer, more distinctive string, require a word boundary, or lean on the domain check — which is naturally unambiguous.
• Country and language defaults matter. Requests default to country: "US" and language: "en". If you sell in Germany, a US-English answer tells you nothing about your visibility there. Set them per market and keep the results in separate series.
• Slow surfaces can come back `failed`. Requests wait up to 180 seconds, and some chatbot scrapes exceed that budget: the record returns failed with a providerFields.snapshot_id and costs zero credits. Retry with that id and the same single surface to redeem the finished answer instead of paying for a re-scrape.

Summary

For a one-off, open ChatGPT and read the answer — it's free and it's fast. The moment the check has to repeat, cover more than a couple of phrasings, or produce something you can compare with last month, fetch the answer as data and test the two signals in code you own. Mentioned and cited are not the same result, and knowing which one you have is what tells you what to fix.