How to Tell If ChatGPT Is Recommending Your Competitor Instead of You

Somewhere in the last year, a chunk of your buyers stopped Googling and started asking. "Best project management tool for a 10-person agency." "Alternative to X that doesn't lock me into annual billing." "Who actually has good customer support in this category." They're typing these straight into ChatGPT, or asking Gemini inside a Google search, or pasting a question into Claude before they've opened a single vendor website.

You don't see any of this happening. There's no impression count, no search-console graph, no rank position to check obsessively. The AI just answers, the buyer reads three names, and one of them isn't yours. You never find out, until a deal falls through and you ask why, and the answer is some version of "we went with the one ChatGPT mentioned first."

The check: five questions, five minutes

You don't need a tool to get a first read on this. Open ChatGPT, Claude, Gemini, Perplexity and Copilot in five tabs and ask each one the same question, phrased the way a buyer would actually ask it, not the way you'd phrase it internally:

• "What's the best [your category] for [your ICP]?"
• "[Your category] alternatives to [your biggest competitor]"
• "Who has the best [the thing you're actually known for]?"

Read the answers looking for three things, not one:

1. Are you named at all? This is the bar everyone checks and stops at. It's the least useful of the three.
2. Are you named first, or fourth? Position inside the answer matters roughly the way SERP position used to, most people don't read past the first two or three names.
3. Is what's said about you accurate? This is the one people skip entirely, and it's often the most damaging. A model that describes your pricing wrong, or attributes a feature to you that you don't have, or, worse, describes a feature you deprecated a year ago, is actively costing you deals with buyers who never talk to you to find out otherwise.

Why "we're mentioned" isn't the win condition

Being named and being cited are different things, and the difference is where most of the actual opportunity is. A citation means the model is pointing at a specific page of yours as the source for a specific claim, pricing, a feature, a comparison. A mention is just your name appearing in a list with no page backing it. Citations are what get built into an answer's trust; mentions are decoration.

If you run the check above and find you're mentioned but never cited, named, but the model isn't pointing at anything you actually published, that's usually not a visibility problem. It's a content-structure problem. The model has heard of you from somewhere (review sites, forums, your own marketing), but nothing on your site is shaped in a way an AI system can confidently point to as a direct answer.

The fix for "not cited" almost never means writing more content. It means restructuring the content you already have so a model can extract a clean, attributable answer from it.

What actually moves the needle

1. Answer the exact question, in the first two sentences

If your pricing page buries the actual number three paragraphs into marketing copy, a model can't confidently cite it. Put the direct answer first, elaborate after.

2. Structure comparisons explicitly

If buyers are asking "X vs Y," and you don't have a page that directly compares you to your top two alternatives, you're leaving that exact query to whoever does have one, usually the competitor, or a third-party review site neither of you controls.

3. Keep claims current

Models are citing a snapshot of your site from whenever it was last crawled and indexed by whatever pulled the training or retrieval data. A pricing page, feature list, or plan comparison that's stale doesn't just risk a wrong citation, it risks a citation that actively argues against you.

4. Check regularly, not once

Model answers shift as they get updated and as your content (and your competitors') changes. A one-time check tells you where you stood on the day you checked. The real value is in catching the moment a competitor starts getting cited for a query you used to own, while there's still time to do something about it.

The five-minute manual version above is worth doing today, once, so you know where you stand right now. Whether it's worth doing every week by hand is a different question, and for most teams, the answer is that it isn't, which is exactly the gap a scheduled scan is built to close.