You built a good product. Customers renew. Then a buyer tells you they asked ChatGPT for options in your category and your name was not in the answer.
That outcome is diagnosable. Models skip brands for four recurring reasons, and a week of structured checking tells you which one is yours.
The Direct Answer
ChatGPT does not recommend you because its process never reaches you. The chain is retrieve, evaluate, synthesize. If retrieval finds nothing credible about your brand, evaluation has nothing to judge, and synthesis names other companies.
Product quality is not in the chain. Evidence quality is. Below are the four evidence failures, ordered by how often they turn out to be the cause.
Reason 1: Weak Entity Signals
The model needs to know exactly who you are before it can mention you. Entities are built from consistency: identical company name, product name, category, and description across your site, Crunchbase, LinkedIn, G2, and GitHub.
Failure looks like this: your Crunchbase says one category, your LinkedIn another, and a 2024 press piece describes the product you replaced. The model cannot resolve the conflict, so it defaults to better documented competitors.
Check it: search your brand on each profile. Any mismatch in category wording, product names, or ownership is a finding. Fix the mismatches before touching anything else.
Reason 2: No Citation Trail
Your website is a claim. Models act on corroboration: reviews, directories, comparison articles, community threads, press. With no independent trail, your claims are unverifiable and unverifiable brands lose ties.
Failure looks like this: searching your brand plus your category returns only pages you wrote.
Check it: run your top category prompt and list every source the answer cites. Are you present in even half of them? Each absent source is a named, reachable target. Start with review platforms and one comparison placement, then keep a steady cadence rather than a burst.
Reason 3: Unstructured Content
Models extract facts from structure: headings that state facts, bulleted features, explicit pricing, schema markup, stable pages rendered server-side. A beautiful scroll narrative teaches them nothing.
Failure looks like this: your homepage communicates through metaphor and your value proposition arrives in paragraph four.
Check it: open your top five pages and ask what a machine learns from headings alone. Then view source. If key content loads only through client JavaScript or schema is absent, you have failed the machine readability test regardless of how good the writing is.
Reason 4: Unclear Positioning
"Leading platform for modern teams" gives the model nothing to file. It needs a category, a buyer, and a differentiator stated plainly: contract analytics for legal teams, incident management for hospitals.
Failure looks like this: the model describes you in the wrong category, merges you with a competitor, or omits the feature that wins deals.
Check it: ask ChatGPT "what does [your company] do" in a fresh session. Compare the answer to your intended positioning word for word. Every wrong noun is positioning debt showing up where buyers now look first.
The Diagnosis Sequence
Run this week, in order:
- Ask your ten highest-value buyer prompts across ChatGPT, Claude, Gemini, and Perplexity. Record who appears and which sources get cited.
- Audit your entity profiles for mismatches. Fix every one.
- Check your presence in the sources those answers cited. List the absences as targets.
- Test your five key pages for machine readability: headings alone, view source, schema validator.
- Ask what the model thinks you do and diff it against intended positioning.
Whichever reason produced the most findings is your first workstream. The free Analyst on this site compresses step one into a single scored run if you want the baseline in minutes instead of days.
