ChatGPT doesn’t recommend the brand mentioned most. It recommends the brand whose story matches the frame it already learned about your category.
Mentions Are Vanity. Narrative Frames Win.
Ask ChatGPT “what’s the best project management tool for a remote team” and watch what happens. It doesn’t run a live tally of who’s mentioned most across the web this week. It reaches for a frame it already holds about that category: what problem it solves, which brands live in which role, what the tradeoffs are.
If your brand shows up in hundreds of listicles but never fits cleanly into that frame, you get mentioned and skipped in the same breath. That’s the gap most teams miss when they check “are we in ChatGPT” and stop there. Being named isn’t the same as being recommended. The model can know your name and still tell a different story about who solves the problem.
This is why mention tracking feels productive and delivers nothing. You can see your brand pop up in an answer, screenshot it, put it in a deck. But the frame underneath the answer, the story about what your category is and who belongs where in it, is the thing actually deciding whether you get the recommendation or a footnote.
How ChatGPT Builds Its Answer: The Three Layers
There’s a structure to how these answers get built, and it’s not a single lookup.
Training frame. The model learned a consensus story about your category from everything it read: articles, comparisons, forum threads, docs. That story hardened into a default frame long before your latest press release existed.
Source bias. When the model reaches for supporting detail, it leans on sources it’s learned to trust for that category: certain review sites, certain publications, certain comparison pages. If those sources never described you the way you describe yourself, the model won’t either.
Recommendation logic. The model matches a query to a role in its frame (“best for small teams,” “most affordable,” “enterprise-grade”) and picks whichever brand fits that role most cleanly. This is why you can be excellent and still lose the slot: you fit a role the model isn’t currently asking about.
Mention frequency barely touches any of these three layers. You can be everywhere in the source material and still be absent from the frame if nothing ever positioned you inside it.
Two Brands, Same Mention Count, Different Recommendation
Picture two project management tools. Call them Brand A and Brand B. Both get cited across roughly the same number of articles, review sites, and comparison pages. Mention share: basically tied.
Ask ChatGPT to recommend a tool for a 10-person startup, and it names Brand B first every time. Why? Brand B’s mentions consistently frame it as “simple, fast to set up, built for small teams.” Brand A’s mentions are scattered: some call it enterprise-grade, some call it a Slack alternative, some just list it without context.
Brand A has the mentions. Brand B has the frame. The model isn’t rewarding volume, it’s rewarding a consistent, learnable story it can slot into an answer. This is the exact scenario behind “why am I mentioned in ChatGPT but never recommended”: your name is in the training data, but it’s not attached to a clear role the model can retrieve.
Where Your Mentions Live (but Your Narrative Doesn’t)
A lot of brands rack up mentions in places that don’t shape the frame at all: a passing line in a “top 20 tools” roundup, a mention in a Reddit thread that got buried, a directory listing with no context. These count in a mention tracker. They do almost nothing for the model’s underlying story about you.
Meanwhile, the handful of sources that actually get cited and re-cited, the ones the model learned to trust, are quietly writing your positioning for you, whether or not it matches what you’d choose. If those sources describe you as a budget option when you’re trying to sell enterprise, that’s the frame ChatGPT will keep using, no matter how many other mentions pile up elsewhere.
Narrative Share vs. Mention Share: Why the Gap Matters
Mention share tells you how often you show up. Narrative share tells you whose version of the story the model actually adopted when it gave the recommendation. These two numbers can move in completely different directions.
You can grow mention share for a quarter, through PR, guest posts, review site outreach, and see zero movement in how often you get recommended. That’s because mention share measures presence. Narrative share measures whether the frame the model uses matches the frame you want. One is countable. The other requires figuring out which sources are actually shaping the model’s story and whether that story is the one you’d sign off on.
The Real Question ChatGPT Answers (Not the One You’re Asking)
When someone asks ChatGPT to recommend a tool, they’re not really asking “who’s mentioned most.” They’re asking “which brand fits the role I need filled.” ChatGPT answers that second question using the frame it learned, drawn from the sources it trusts, applied to the specific role in the query.
That means the useful question for your team isn’t “are we mentioned in ChatGPT.” It’s “what frame does ChatGPT use for our category, whose version of that frame is winning, and which sources are teaching it that story.” Those are the things that actually move a recommendation. Mention count is just the noise sitting on top.
If you want to know whose frame ChatGPT is actually running with, and why it’s picking your competitor over you, that’s the analysis Mavel does. Come talk to us before you spend another quarter chasing mentions that were never going to move the answer.
