Being named in an AI answer doesn’t mean you won it. Here’s how to check whose story the model is actually telling about your category.
Why Mention Counts Lie (The Mentions ≠ Narrative Problem)
Ask ChatGPT “what’s the best project management tool for small teams?” and you’ll probably see your brand somewhere in the answer. That feels like a win. Someone on your team screenshots it and drops it in Slack.
But look closer at how you show up. Are you the third option in a list, described in one flat sentence, while a competitor gets two paragraphs about why they’re the “most flexible choice for growing teams”? That’s not a tie. That’s a loss dressed up as a mention.
This is the trap with most AI-visibility tracking today. Tools count appearances: how many times your name shows up across a set of prompts, in what position, alongside which competitors. That’s useful data. It’s also incomplete in a way that can mislead you badly.
A brand can appear in 50 AI answers and still be losing the category, because the model has learned a story where that brand is the safe-but-boring option and a competitor is the innovative one. Mention counts don’t capture that. They just count names. The actual thing deciding who gets recommended is the frame: the interpretive story the model has learned about your category and where each brand fits in it.
The Narrative Audit: What AI Actually Believes About Your Brand
A narrative audit starts from a different question than a visibility check. Instead of “where do we show up,” you ask: “what does the model believe is true about us, and where did it learn that?”
Run this test yourself. Open ChatGPT, Perplexity, and Google AI Overviews and ask each one:
- “What is [your category]?”
- “What’s the best [your product type]?”
- “Is [your brand] good for [a specific use case]?”
- “What do experts say about [your brand]?”
Don’t just note whether you appear. Write down the actual adjectives and framing each model uses. Are you “enterprise-grade” or “budget-friendly”? “Complex but powerful” or “simple but limited”? That language is the narrative. It’s what the model has concluded about you from everything it’s absorbed, and it’s usually more consistent, and harder to shake, than any single answer suggests.
Read the Frame, Not the List (Where to Find the Narrative in AI Answers)
Most people scan AI answers for their brand name and stop reading. The narrative audit means reading the whole answer as a story with a plot, not a list with entries.
Ask: “Compare [your brand] vs [competitor].” Read how the model sets up the comparison. Does it frame the category around a problem your competitor solves better by definition? If the model opens with “when choosing a CRM, the main tradeoff is ease of use versus customization,” and then slots your competitor into “easy” and you into “customizable,” that framing was decided before your name ever appeared. You’re not losing the comparison. You’re losing the frame the comparison runs on.
Try the direct version too: “Why do people recommend [competitor] over [your brand]?” The answer will often reveal the model’s assumed narrative more clearly than any neutral prompt, because it’s forced to justify a preference. Pay attention to what it cites as reasons. That’s your gap list, right there in the response.
Map Your Narrative vs. Competitor Frames (The Consensus Question)
AI models don’t invent opinions from nothing. They’re downstream of consensus: the aggregate of reviews, comparison articles, forum threads, and analyst posts that already exist about your category. If that consensus leans a certain way, the model will reflect it, and sometimes amplify it.
So the audit needs a side-by-side. For each major competitor, run the same set of prompts and log the frame the model uses for them versus the frame it uses for you. Put it in a simple table: category definition, primary strength claimed, primary weakness implied, use cases it’s recommended for. Patterns show up fast. You might find every competitor gets described in terms of outcomes (“helps teams ship faster”) while you get described in terms of features (“has a drag-and-drop builder”). That’s not a coincidence. That’s whose narrative is winning the category-level story, independent of who gets mentioned more often.
The Source Intelligence Layer (Why Citations Matter More Than Presence)
Once you see a frame you don’t like, the next question is where it came from. Models cite sources, or their answers clearly draw on identifiable content: G2 comparison pages, Reddit threads, review roundups, analyst blog posts. Trace those back.
If AI recommendation bias toward a competitor keeps showing up, check whether it’s because three widely-cited comparison articles all use the same framing, possibly copying each other. That’s fixable. A dashboard telling you “you’re mentioned less positively” doesn’t tell you that. Source intelligence does, because it points at the actual input shaping the output, not just a score describing the symptom.
Build Your Audit Checklist (Actionable Narrative Gaps to Fix)
Turn what you’ve found into a short, prioritized list instead of a research document nobody reopens:
- Which prompts return a frame that misrepresents you, and how
- Which competitor frame is winning the category definition itself
- Which sources are most cited in comparisons where you lose the frame
- Which use cases you’re absent from entirely, not just under-represented in
- What’s the single narrative shift, one sentence, that would change the most answers
That last one matters most. A narrative audit that ends in a spreadsheet of mentions is just visibility tracking with extra steps. The point is to find the one or two frame corrections worth shipping content, PR, or product messaging against.
Want someone to actually run this audit on your category instead of doing it by hand across five tools? That’s what Mavel does. Come see whose frame the model is really using for your brand.
