A B2B SaaS that never appears in ChatGPT answers usually has one of 6 problems, and none of them are mysterious.
- ChatGPT pulls 74.6% of product-query citations from vendors' own sites. Perplexity, Gemini, and Claude pull 79% from third-party sources. Either layer can fail on its own.
- The 6 causes: blocked crawlers, category ambiguity, thin third-party evidence, missing decision-stage pages, stale information, and competitors owning your comparisons.
- A 20-minute self-check tells you which ones are yours.
- Every cause maps to a fix, and most fixes start moving within a quarter.
Buyers ask the engines for the best option in a category, the engines answer, and some brands never come up. When it’s your brand, the silence feels personal, but across the audits we run, the same 6 causes explain nearly every absence, and each one has a fix you can start this quarter.
How ChatGPT decides who appears
ChatGPT builds product answers from 2 places: your own website, which it visits directly, and what the rest of the internet says about you. Our citation study found GPT-5.4 pulls 74.6% of its product-query citations from vendors’ own sites, while Perplexity, Gemini, and Claude pull 79.0% from third-party sources.
Those 2 layers fail independently, which is why so many teams are confused about their absence. A site can hold page 1 rankings on Google while the review sites, listicles, and comparison pages that feed the other engines say nothing about it. And a brand with glowing third-party coverage can still miss on ChatGPT because its own pricing page is unreadable to a crawler. Before you fix anything, you need to know which layer is failing, and for most companies the answer is one of the 6 causes below.
The 6 causes, in the order we find them
Blocked crawlers, category ambiguity, thin third-party evidence, missing decision-stage pages, stale information, and competitor-owned comparisons. Most companies carry 2 or 3 of these at once, so read all 6 before deciding which fix comes first.
Cause 1: The crawlers can’t get in
The engines send their own bots: GPTBot for ChatGPT, ClaudeBot, PerplexityBot, and Google’s crawlers for AI Overviews. A robots.txt rule, a CDN bot-protection setting, or an overzealous firewall can lock them all out, and the engine treats a page it can’t read as a page that doesn’t exist. Check your robots.txt and server logs first, because this is the cheapest fix on the list and it gates every other one.
Cause 2: The engines can’t tell what you are
If your homepage says “revenue platform,” your LinkedIn says “sales intelligence,” and G2 lists you under something else again, the engine can’t place you in a category with confidence, so it recommends a brand it can place. This is why we tell clients to build the entity first: one description of what you are and who you’re for, used identically across your site, your profiles, and the sources that describe you, until the engines stop hedging.
Cause 3: Nobody else says anything about you
This is the most common cause we find. With 79% of product-query citations pointing at third-party sources on 3 of the 4 major engines, a brand with no reviews, no listicle placements, and no independent comparisons has starved the engines of evidence. You can’t fix this on your own website. It’s just how the game works: the engines want corroboration, and corroboration only comes from other people’s pages.
Cause 4: You have no decision-stage pages
The engines cite different formats at different buying stages. In our citation data, listicles dominate consideration queries at 50%, comparison pages take evaluation at 42%, and pricing content wins decision queries. A blog full of “what is X” explainers gives the engines nothing to cite when a buyer asks which product to pick, because that question gets answered from comparisons and pricing pages you may never have built. And the pages you do build need to be quotable: plain headings, real numbers, tables over images, sentences an engine can lift whole.
Cause 5: What’s out there about you is stale or wrong
Engines repeat what they read. If the most-cited comparison of your product quotes pricing from 2 years ago or describes a feature you renamed, that’s what buyers hear. Stale information does something worse than keeping you out of the answers: it puts a wrong version of you in them, and a wrong version can cost more than absence.
Cause 6: Your competitors wrote your comparisons
When the best-ranking “you vs competitor” page was published by the competitor, every buyer who asks the engines for that comparison gets their framing. We tend to treat this one as the loudest alarm in an audit, because it means the evidence layer has gone past thin and started working against you. The reverse also holds: comparison prompts nobody has claimed are open territory, and they’re usually the cheapest wins available.
The 20-minute check that tells you which
Run 10 buying prompts for your category across ChatGPT and Perplexity, and classify every appearance of your brand as Recommended, Cited, Mentioned, or Absent. Then check robots.txt for blocked bots, and search for the top comparison pages naming you. Those 3 checks expose which of the 6 causes you’re carrying.
That’s the lean version. The full process, with the query sets, the logging sheet, and the scoring, is in our step-by-step audit guide, and it takes an afternoon.
What fixes what
Each cause maps to a specific fix, and the mapping matters, because we feel as though half the wasted quarters in this discipline come from fixing the wrong layer: rewriting their website when the evidence layer is the problem, or chasing reviews when the crawlers were blocked all along.
| Cause | The fix |
|---|---|
| Crawlers blocked | Allow the AI bots in robots.txt and your CDN, then verify with server logs |
| Category ambiguity | One positioning line, used identically across your site, profiles, and directories |
| Thin third-party evidence | Review campaigns on the platform AI cites in your category, plus digital PR for mentions and links |
| No decision-stage pages | Build the comparisons, alternatives pages, and pricing content buyers ask for |
| Stale information | Refresh your own pages and request corrections from the sources engines cite |
| Competitor-owned comparisons | Publish your own comparison pages and earn independent ones from third parties |
How long until you show up
Crawler fixes register in days. New decision-stage pages start getting retrieved in weeks. Evidence-layer work, reviews and mentions and independent comparisons, compounds over months. In our experience the first visible movement lands inside a quarter, and it accelerates from there.
I’d say the discipline that matters most is measurement, because the engines update quietly. Track the same prompts every month, watch your tier move from Absent to Mentioned to Recommended, and resist judging the whole effort in week 2. If you’re a lean team, you have to pick your top 2 causes and ignore the rest for a quarter, because half-fixing all 6 moves nothing. The answers move when the evidence moves, and the evidence takes time to build.
Frequently asked questions
The 3 questions we hear most from teams who’ve just discovered their brand missing from AI answers, answered the way we answer them on calls: why competitors appear, whether Google rankings carry over, and whether every engine behaves the same way.
Why do competitors show up in ChatGPT when we don’t?
Because they have more evidence. More reviews on the platforms AI cites, more listicle placements, more independent comparisons. The engines weigh what’s written rather than ranking product quality, and whoever has more written about them wins the answer.
Does ranking on Google help you appear in ChatGPT?
It helps, though I do think teams overrate the carryover. The engines retrieve from Google’s index, so your rankings feed the pipeline, but ChatGPT also goes directly to vendor sites and the other engines lean on third-party sources your rank tracker never measures. Treat Google as necessary, and the evidence layer as the part your SEO never covered.
Is it the same on Perplexity and Gemini?
The causes are the same, the weighting differs. Perplexity and Gemini lean harder on third-party content, so thin evidence hurts you more there. ChatGPT leans on your own site, so crawler access and page clarity matter more. Fixing all 6 causes covers every engine at once.
