Visible in ChatGPT: Check Your Company's AI Visibility

In a few minutes you know whether AI assistants like ChatGPT or Perplexity recommend your company when someone in your industry asks for a recommendation. And if not: who gets named instead and where those answers come from. No write access, no change to your accounts, a pure status check.

Someone asks an AI assistant: "Which [industry] in [place] can you recommend?" The answer names a few, and if yours isn't among them, you lose those customers without noticing. AI visibility (often called GEO, for Generative Engine Optimization) means exactly that: whether and how you appear in the answers of ChatGPT, Perplexity, and Google's AI overview. By the end of this tutorial you have three things: a number for how often your company shows up in AI answers, a list of who gets recommended instead, and the sources the AIs draw their answers from. That's the foundation for deciding whether and where work is worth it.

One point up front so you read the result correctly: your agent doesn't ask ChatGPT live. It taps a large, collected index of millions of AI answers and counts within it where your company and your competitors appear. That's a sample, not a live measurement. More on this below under "Honest limit".

What you need

• DataForSEO connected (required). The whole measurement runs through this source: the AI queries tap DataForSEO's collected answer index (billed by usage, a few cents per query from your credit). Without DataForSEO, your agent has no AI-visibility data. To connect it, see Connect Marketing Accounts to Claude; for ChatGPT, Connect Marketing Accounts to ChatGPT shows the same setup.
• Search Console connected (optional). With it you additionally check what Google already knows you for today (your own search terms). Helpful context, but not needed for the four steps. Without Search Console, you skip the optional extra check in the variations.

1. Status check: is your company named?

First the sober question: does your company appear in AI answers at all?

Check for [workspace] how often my company "client-a" is named
and cited with a link in AI answers about my industry.
Show the numbers separately by AI assistant.

What you get back:

Mentions of "client-a" in AI answers:        1
  Google (AI overview):                      1
  ChatGPT:                                   no data in the index
Citations with a link to your site:          0

For many smaller companies there's a 0 or a very low number here at the start. That isn't an error, but exactly the starting point: you're barely recommended yet. What matters is the difference between "named" (your name comes up) and "cited" (the AI actively links your page). Cited is the stronger signal.

✓ Checkpoint: You have a number for how often your company is named and how often it's cited, even if it's 0.

Doesn't work? If you get "subscription_required" or an access error, either DataForSEO isn't connected (see "What you need") or the credit is used up. If a 0 comes back for an AI assistant, that often just means: too few collected answers on your niche topic, not necessarily "invisible there".

2. Who gets recommended instead?

A 0 for you is only half the story. Now you turn the question around: when someone in your industry asks for a recommendation, who do the AIs name? Sorted by citations, the stronger signal from step 1.

When someone asks an AI assistant for the best [industry] in
[region]: which providers and which websites get named most
often? And where do I stand in comparison?

What you get back:

For "best marketing agency [region]" the AIs name most often:

  business-directory.at     18 citations  (directory)
  agency-north.at           17 citations  (competitor)
  webstrong.at              15 citations  (competitor)
  marketing-guide.de        12 citations  (comparison blog)
  review-portal.at           9 citations  (review site)

  client-a:                  not in the answers

Now you see the competitive picture: direct competitors, but also directories, review sites, and comparison blogs. These are exactly the pages that help determine who the AI recommends.

✓ Checkpoint: You have a ranking of the most frequently named providers and pages, and you see at which position (or whether at all) your company appears.

Doesn't work? If a very short or empty list comes back, the question is too niche or too local. Make the category one step broader: region instead of neighborhood, umbrella term instead of specialized service.

3. Where the AIs draw their answers from

Now the decisive question for your strategy: which concrete sources do the AIs build these recommendations from? Because an AI can't point to you on its own. It cites pages it trusts.

Show me for the topic [industry/region] which pages the AIs
concretely cite, and for which question each time. Group by
type of source (directory, review, competitor, guide).

What you get back:

AI question: "digital marketing agency"
  cites: business-directory.at   → entry "Online agency"
  cites: agency-north.at         → homepage
  cites: marketing-guide.de      → "What does an agency cost?"

AI question: "online marketing provider [region]"
  cites: review-portal.at        → provider comparison
  cites: webstrong.at            → services page

You learn two things here. First, the AI breaks your topic down into real customer questions (what does it cost, which provider, which comparison). Second, almost all citations come from third-party pages, not from the companies themselves. Brands are cited far more often via outside pages (directories, reviews, comparisons) than via their own website. So your own site alone rarely gets you into the answer. You have to show up where the AI is already looking anyway.

✓ Checkpoint: You have a list of concrete source pages with the respective question, and you recognize the pattern of which type of pages (directory, review, guide) dominates the answers.

Doesn't work? If the list shows only definition pages (Wikipedia, encyclopedias), the question was too general. Add a buying intent: "best", "provider", "cost", "comparison".

4. Derive the levers (honestly)

The status check now becomes a plan. Have concrete actions derived from the three steps, and be honest with yourself about what that means in effort.

Summarize the three results: what concretely do I have to do so
that I show up in the AI answers for these questions? Prioritize by
effort and impact, and say honestly how long it takes.

What you get back:

1. Be present where the AI already cites:
   listings/profiles on business-directory.at, review-portal.at.
   Effort: low to medium. Impact: high.
2. Answer the real customer questions on your site:
   "What does [service] cost?" with clear numbers, a comparison table,
   the answer right in the first paragraph. Effort: medium. Impact: medium.
3. Build third-party mentions: collect reviews, appear in
   comparison articles and lists. Effort: high. Impact: high.

The most important point first, without sugarcoating: this is content and PR work, not a switch. A directory listing is quickly done, but until AIs name you more often, weeks to months are more likely to pass. The answer indexes update slowly, and trust on third-party pages builds slowly. So measure monthly (repeat steps 1 and 2), not daily.

✓ Checkpoint: You have a list of concrete actions sorted by effort and impact, and a realistic expectation: impact comes in weeks to months, not in days.

Doesn't work? If the suggestions stay vague ("more content"), give your agent the concrete source list from step 3: "Derive the actions from exactly these cited pages."

Variations

• "Check the same questions in English." (English questions pull international AI answers, relevant if you want to be visible beyond the DACH region.)
• "Compare my AI visibility with [competitor]: where do they get named and I don't?" (pins the gap to a concrete name)
• "What does Google already know me for today? Show my top search terms, split into brand and topic searches." (uses Search Console, only with Search Console connected, otherwise skip)

Honest limit

• AI answers fluctuate. The same question can name different companies today and tomorrow: personalized, depending on the model version and session. From a single query you must not derive a "score", only a trend from repeated measurement.
• It's a sample, not a live query. Your agent reads a collected index that isn't up to the day. Freshly published content appears only with a delay.
• Not every AI assistant provides data. For niche or local DACH topics, often only one assistant (frequently Google's AI overview) has enough answers in the index. A 0 for ChatGPT can mean "too little data", not "invisible there".
• No guarantee. No one can promise you that after the work you'll "appear in ChatGPT". You improve the probability, nothing more. The models decide for themselves what they cite.
• Pure diagnosis. This tutorial changes nothing in your accounts or on your website, it only measures. The implementation (listings, content, reviews) is a separate, manual step.

What's next

You now know which questions the AIs ask and which answers they reward: direct, well-supported answers to real customer questions. Exactly such an article you write in the next tutorial and publish it right away. → Write a Blog Post and Publish It