ChatGPT may mention a competitor because the question fits that company, relevant information is easier to retrieve, the model is familiar with the brand or the answer varies between runs. One missing mention does not identify the cause. Start by examining the question and the evidence behind the answer.
Key points
- The investigation below tests four explanations rather than assuming every missing mention is a content problem.
- A brand can be mentioned without a supporting citation; a cited article need not imply a recommendation.
- Google’s guidance on query fan-out applies to Google AI search, not automatically to every ChatGPT response.
- The sample investigation is hypothetical and contains no claimed client visibility results.
Is the question a fair test of your business?
Check that your company genuinely fits the buyer’s constraints. A prompt asking for a large agency with in-house production is not a fair test of an independent consultant. A request for low-cost self-service software may not fit an enterprise service.
Write the buyer, problem, market and constraints before testing. Include both open discovery prompts and realistic evaluation questions. Asking only about your own brand measures something different from seeing whether it appears without prompting.
What explanations should you compare?
Keep several explanations open until the evidence narrows them.
| Explanation | What to inspect | What would weaken it |
|---|---|---|
| Your offer does not fit | The prompt’s constraints against actual capabilities | A clearly documented match |
| Relevant information is difficult to find | Accessible pages, search results and cited sources | The answer already retrieves your relevant page |
| Evidence is unclear or inconsistent | Descriptions of audience, scope and proof | Consistent supported descriptions across relevant sources |
| The observation is unstable | Repeated runs with the same recorded conditions | A persistent pattern across the agreed sample |
This is a diagnostic framework, not a ranking model. It cannot reveal the internal reason for a particular generated answer.
How do you inspect the sources?
Open the cited pages and identify what they actually support. Record whether each source is a provider page, review, directory, editorial comparison or other material. Note whether it supports the stated capability, merely names the business or is being interpreted too broadly.
Do not assume inclusion in one directory caused the recommendation. A competitor may have several relevant sources or stronger fit. Look for factual gaps you can address honestly, not instructions to manufacture mentions.
What would a practical investigation look like?
Start with a narrow question and one change whose purpose you can explain. In this hypothetical example, a specialist consultant is omitted from answers about help with poor-quality SaaS Google Ads leads. The site describes campaign management but does not clearly explain qualification analysis.
The next step is to verify that this work is actually offered, then document the process, data required and limitations on the relevant service page. Add an example diagnosis and link it from the supporting article. Preserve the original prompt log and repeat the observations later.
If mentions increase afterwards, report the timing and other changes. Do not claim causation from a before-and-after screenshot. If nothing changes, the clearer page may still help human buyers, but that is a separate result to measure.
What should you avoid?
Avoid actions that manufacture confidence rather than improve evidence.
- Do not publish fake reviews or unsupported ‘best consultant’ claims.
- Do not copy competitor wording or invent capabilities.
- Do not count every mention as a qualified lead.
- Do not change the prompt sample until the report looks favourable.
- Do not mistake a new heading or schema block for a proven citation mechanism.
For a wider review, see my AI visibility approach.
Frequently asked questions
Does ranking on Google guarantee a ChatGPT mention?
No. Search systems, query wording and generated answers differ. A Google position alone does not establish what another platform will retrieve or recommend.
Should I add more articles?
Only when a relevant buyer question lacks a useful answer. An existing commercial page may need clearer evidence before another broad article is justified.
Should I compare my company with competitors?
A fair comparison can help buyers if it uses verified criteria and explains trade-offs. Do not present your commercial perspective as an independent ranking.
Can I buy guaranteed recommendations?
I do not offer guaranteed recommendations or citations. Any proposal should explain what is actually being delivered and how results are observed.
Where I can help
I work directly with B2B and SaaS companies on Google Ads and SEO and AI visibility. A Search Diagnostic can help when the main constraint is unclear. If your team already knows what needs changing and can implement it, a separate diagnostic may not be necessary.
Book a 30-minute call. Tell me what you want to improve and we can work out whether there is a useful fit.
Sources and methodology
Recommendations are a practical editorial and diagnostic approach, not a proven formula for AI inclusion. See Google’s guide to generative AI search and AI features and your website. Google guidance should not be generalised into guarantees about other platforms. My published paid-search experience should not be interpreted as historical SEO or AI-citation results.
Sources checked: 4 November 2026. Examples are illustrative unless explicitly described otherwise.