Does ChatGPT Recommend Your Product? We Measured Ours — and Scored 0%

Why AI answers are a different battleground

Classic SEO optimises for a page of ten blue links where being #7 still earns clicks. An AI assistant does not return ten options — it usually names two to five, in a sentence or a short list. Position four on Google is a bad day; position four in an AI answer often means you simply are not mentioned.

The second difference is attribution. AI answers are assembled from sources the engine trusts — documentation, comparison articles, directories, forums, review sites — and then paraphrased. You can be excellent and still be invisible, because the engine never encountered a source that described you in the context of the question being asked.

That is what "generative engine optimization" (GEO) means in practice: making sure the sources an engine reads contain a clear, checkable description of what you do, for the questions your buyers actually ask.

How to measure it honestly

The measurement is simple to describe and easy to get wrong. You ask each engine a fixed set of buyer questions, then count how often a company is named. Four details separate a real number from a flattering one:

• Use the questions buyers actually type, not your brand name. "Best competitive intelligence software for startups" is a buying question. "What is ClientCues?" is a lookup.
• Separate branded from category questions. Any question containing your name will almost always mention you. Mixing the two inflates your score and hides the real problem.
• Ask more than one engine. ChatGPT, Gemini and Claude read different sources and disagree constantly. A single-engine reading is anecdote, not measurement.
• Count citations separately from mentions. Being named is good; having your domain cited as the source is better, because it survives the answer being rephrased.

What we found when we measured ourselves

We ran 18 buyer questions across three engines. The split between branded and category questions is where the story is:

Had we reported a blended number, we could have honestly written "17% AI visibility" and felt fine about it. Split apart, the blended number is meaningless: we win the questions that already contain our name, and lose every question a new buyer would actually ask.

The competitive picture on those same category questions:

Klue is not winning because its product is ten times better. It is winning because years of review-site presence, comparison articles and directory listings taught the engines that "Klue" is an answer to "competitive intelligence software". That is a content and citation gap — and unlike a product gap, it is closable by writing and being listed.

Check your own visibility in ten minutes

You do not need a tool for a first reading. You need a method you can repeat.

• Write five category questions a buyer would ask before they know any vendor names. No brand names allowed.
• Ask each one in ChatGPT, Gemini and Claude with web search enabled, in a fresh chat — memory and chat history will bias the answer toward you.
• Record three things per answer: were you named, in what position, and which sources were cited.
• Repeat monthly. A single reading is noise; the trend is the signal. Answers vary between runs, so treat any one result as approximate.
• Read the citations, not just the answer. Those cited pages are your real competition for the next answer — they are where you need to be present, or alongside.

What actually moves the number

Two things decide whether an engine names you: whether trusted sources describe you in the language of the question, and whether your own pages are structured enough to extract cleanly.

• Comparison and alternatives content. Engines quote pages that lay out options side by side, including honest weaknesses. A page that only says you are the best is a page they cannot safely cite.
• Be present where engines already look. Review directories, category listings and community threads are cited disproportionately often. Absence from them is the most common reason a good product is invisible.
• Answer the question literally on the page. A heading phrased exactly like the buyer question, followed by a direct two-sentence answer, is far more extractable than the same information spread across five paragraphs.
• Add structured data. FAQPage and Article schema make question-and-answer pairs machine-readable — the format assistants extract most reliably.
• Keep facts checkable and current. Prices, tier names and dates that contradict your own site make an engine less likely to quote you at all.
• Give it weeks, then re-measure. Sources are re-crawled on their own schedule; changes show up over weeks, not days.

The honest caveats

AI answers are non-deterministic. Ask the same question twice and different companies may be named — which is exactly why one check tells you little and a tracked trend tells you a lot.

Engines also differ sharply. In our own runs, coverage varied by engine for the same question set, so treat any cross-engine average as a rough composite rather than a precise score.

And visibility is not the goal in itself. Being named only matters if the buyer reading the answer fits your product: 40% presence in the wrong category is worth less than 10% in the right one.

Frequently asked questions

How do I know if ChatGPT recommends my product? Ask it the buying questions in your category — without your brand name — in a fresh chat with search enabled, and record whether you are named and which sources it cites. Repeat monthly, because answers vary between runs. ClientCues automates the same loop across ChatGPT, Gemini and Claude and tracks the trend.

What is generative engine optimization (GEO)? GEO is the practice of getting your product accurately represented in AI-generated answers: being described in the sources engines read, answering buyer questions directly on your own pages, and marking those answers up so they can be extracted. It complements SEO rather than replacing it.

Why does my brand appear when I ask about my own company but nowhere else? Because a question containing your name retrieves pages about you. Neutral category questions retrieve pages about the category — and if none of those describe you, you will not appear. That gap is the number worth fixing.

How long does it take to improve AI visibility? Weeks, not days. Publishing a comparison page or earning a directory listing today typically shows up in answers over the following weeks, as sources are re-crawled. Measure monthly and judge the trend.

Does traditional SEO still matter? Yes. The pages engines cite are usually pages that also rank, and Google's own AI summaries draw on its index. GEO is an additional lens on the same content, not a replacement for it.