You ask ChatGPT for the best tools in your category. Your competitor is right there in the answer. You are not. No error message, no warning, just absence.
That absence is not random, and it is not bad luck. ChatGPT skips a brand for a small number of specific reasons, and each one leaves a different fingerprint. This guide walks through the five, shows you how to tell which one is yours, and explains what actually closes the gap.
The problem: invisible is worse than ranked low
In classic search, being on page two still meant you existed. Someone could scroll. In an AI answer there is no page two. The model names a few options and stops. If you are not in that short list, you were not beaten, you were never in the room.
And you usually have no idea it is happening. There is no report that says “lost to a ChatGPT recommendation.” The buyer just never arrives, and you never learn why. So the first job is understanding the mechanism, because the fix depends entirely on which reason is keeping you out.
The five reasons AI skips a brand
Each one leaves a different fingerprint, so before the detail, here is the whole set at a glance: how each reason shows up, and what fixes it.
| # | Reason | How to tell | The fix |
|---|---|---|---|
| 1 | Entity weight | ChatGPT can’t clearly say what you are or what you do | Describe yourself the same way everywhere until the model is sure you exist |
| 2 | Co-occurrence | You never appear next to your category in “best X” lists or comparisons | Get into the roundups, comparisons, and buyer discussions in your space |
| 3 | Sentiment | You are mentioned, but always with a caveat or a warning attached | Address the negative narrative third parties carry about you |
| 4 | Source authority | The only place that describes you well is your own website | Earn third-party coverage: reviews, Reddit, G2, editorial, directories |
| 5 | Retrieval | The model can find competitors live, but cannot reach or read your pages | Open crawl access, ungate content, lead each page with a clean answer |
1. Entity weight: the model is not sure you exist
Before ChatGPT can recommend you, it has to be confident you are a real, distinct thing in your category. That confidence is built from how often and how consistently you appear across the web, not from your domain authority. One analysis of thousands of AI citations found that brand search demand was among the strongest signals tied to being cited, while backlinks, the backbone of old SEO, barely correlated. This is why a young brand with sharp, consistent signals can beat an established one with a stronger backlink profile. If the model has only seen you a handful of times, in scattered and inconsistent ways, it treats you as noise and leaves you out.
2. Co-occurrence: you never appear next to the category
Models learn which brands belong to which questions by seeing them together. If your name almost never appears in the same place as your category and your competitors, in comparisons, roundups, “best tools for X” lists, and buyer discussions, then ChatGPT has no pattern that connects you to that prompt. You can be a real, well-described company and still be absent from “best X for Y” simply because nothing ever placed you in that conversation. The brands that get recommended are the ones that keep showing up in evaluation and comparison contexts.
3. Sentiment: when you do appear, the framing is cold
Sentiment is not just a PR concern in AI answers, it is closer to a ranking factor. Models are tuned to be helpful and cautious, so they hesitate to recommend a brand wrapped in consistent complaints or unresolved controversy. A brand that shows up with a hedge attached, “some users report problems with,” is arguably worse off than a brand that is simply not mentioned, because the model is actively steering people away rather than just omitting you.
4. Source authority: only you talk about you
This is the most common culprit, and the most uncomfortable. AI engines lean heavily on independent, third-party sources, and they treat your own website as one of the least objective inputs. Large citation studies keep finding the same thing: the clear majority of AI citations point to external sources, not brand-owned pages, and independent content gets cited several times more often than company sites. Reddit, review platforms, editorial coverage, and category publications carry far more weight than your homepage. If the only place your brand is described well is your own site, the model has nothing trustworthy to lean on, and it picks the competitor that third parties vouch for.
5. Retrieval: the model literally cannot reach you
Modern ChatGPT does not rely on training data alone. For many questions it retrieves live pages from the web and grounds its answer in what it can pull right now. That introduces a purely technical failure mode. If your pages block the retrieval bots, hide content behind logins or forms, render everything in JavaScript, or bury answers in long unstructured paragraphs, the model cannot extract a clean fact to cite, so it reaches for a competitor it can read. Being technically retrievable and easy to extract from is its own requirement, separate from everything above.
How to diagnose which one applies
The reasons look similar from the outside, absence, but they need different fixes. Here is how to tell them apart.
-
Ask the model directly. In ChatGPT, ask “what is [your brand] and what does it do.” If the answer is wrong, vague, or hedged, your problem is entity weight or sentiment, not retrieval. If it describes you accurately but still does not recommend you in category prompts, the issue is co-occurrence or source authority.
-
Run the category prompt, then read the sources. Ask “best [category] tools for [use case]” and look at who gets named and, where the model cites, where it pulled from. If competitors are cited from Reddit, G2, and editorial roundups you are absent from, that is a source-authority and co-occurrence gap.
-
Check the framing, not just the presence. If you are mentioned but always with a caveat, you have a sentiment problem rooted in the narrative third parties have built about you.
-
Test retrieval separately. Confirm the AI search and retrieval bots can actually reach your key pages, that nothing important is gated or rendered only in JavaScript, and that each page leads with a direct, extractable answer. If the model can find competitors live but never you, this is where to look.
What actually moves the needle
Most brands pour effort into the wrong lever. Here is what the evidence supports.
Earn third-party coverage. Because independent sources carry the most weight, the highest-leverage move is getting credibly described somewhere other than your own site, in reviews, comparisons, community threads, and editorial pieces. This single shift addresses source authority and co-occurrence at the same time.
Make your entity consistent. Describe what you are, who you serve, and your category the same way everywhere: your site, your profiles, your press, third-party listings. Models build confidence from repetition of a stable story, and inconsistency lowers it.
Get into the comparison contexts. If you are missing from the “best X” and “X vs Y” content in your space, that is exactly the content to go earn, because that is what the model reads when someone asks for a recommendation.
Fix the technical floor. Make sure retrieval bots can reach you, kill the gating on your best content, and lead each page with a clean, quotable answer. None of the above matters if the model cannot read you.
And here is what is overrated: chasing backlinks alone, stuffing keywords, and polishing your own homepage copy one more time. Those were the old game. They move AI visibility far less than a single credible third-party mention.
Across the brands we scan at Rankry, the pattern is consistent: the ones sitting near zero visibility are almost never failing on one exotic factor. They are usually thin on third-party citations and inconsistent in how they describe themselves, while their site is technically fine.
How to measure it
You cannot fix what you cannot see, and AI visibility is invisible by default. There is no dashboard in ChatGPT telling you which prompts you lost and to whom.
That is the gap worth closing first. Tracking your presence across the prompts your buyers actually ask, on every model, tells you which of the five reasons is really yours, and whether the work you do moves the number. Without it you are guessing, and guessing is how brands spend a year fixing the wrong thing.
If you want the deeper mechanics behind all of this, we break down how LLMs choose which brands to recommend in detail. And if ChatGPT does mention you but never recommends you, that is a different and important distinction we cover in why being mentioned is not the same as being recommended.
