ChatGPT can link your page and leave your brand out because citations and mentions are separate outcomes. The link shows the page was retrieved and used as evidence; the name appears only if the model treats your brand as part of the answer. OpenAI documents no rule for naming, but you can measure and narrow the gap.
What a cited-but-unnamed result means, and how it differs from zero mentions
A citation is a linked source attached to an answer. A brand mention is your name written in the answer text. They come from different steps, so one can happen without the other. Being cited shows your page was retrieved and judged useful for a claim. It does not show that the model treats your brand as part of the answer. See how an AI mention differs from an AI citation.
Hypothetical example, not a client case: a user asks ChatGPT with search on which time-tracking tool suits a small architecture firm. The answer names several tools and explains billable-hour reporting. Your guide to billable hours for architects appears as a source next to that paragraph. Your brand, which sells such a tool, is not among the names. Your page served as evidence; other brands received the recommendation.
Read each answer against three outcomes, which are different problems from never appearing at all:
- Named and cited. Retrieval and entity recognition both worked: the model used your page and treated your brand as part of the answer.
- Cited but not named. Retrieval worked, entity recognition did not. Your content is trusted as evidence, but the model does not link it to your brand or does not see your brand as a candidate answer.
- Named but not cited. Entity recognition worked without retrieval of your page. The model knows your brand, from trained knowledge or from other sources, but did not use your page for that answer.
Never appearing is a separate case with its own causes, covered in diagnosing a brand that gets zero mentions in ChatGPT.
This also answers what showing up in ChatGPT means for a brand. A source link alone is a weaker form of showing up than being named. A reader can act on the answer without opening the sources.
Why ChatGPT links your page but leaves your name out of the answer
What OpenAI documents about search and citations
ChatGPT attaches citations mainly when it retrieves live web results. The page must therefore be reachable by OpenAI's search crawler to be cited at all. OpenAI's crawler documentation, verified on 24 October 2026, states that "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links." A robots.txt rule that blocks OAI-SearchBot removes a page from search answers before naming is even in question.
ChatGPT also answers many prompts from trained knowledge without browsing. Those answers can name a brand with no link attached, which produces the opposite pattern, named without citation, and explains why a brand's mention rate and citation rate can diverge.
Retrieved, cited and named are three distinct stages. A page can be retrieved and cited and still fail the third stage. OpenAI's crawler documentation covers the first. The help center article on ChatGPT search could not be fetched for verification on 24 October 2026, and nothing we could read explains how a brand name enters the answer text.
The likely causes, labelled as inference
Every cause below is inference: no platform documents how names enter answer text, and we have not tested these causes in isolation. They are listed in diagnosis order, from the quickest to check to the hardest to change.
- The passage states facts without naming the subject. If the lifted passage says "our platform" or "it", the fact can travel while the name stays behind. Naming the brand clearly and consistently in the page text is a precondition for the name to travel with the passage.
- The page supports a generic fact. Your page may be cited for a definition, a process or a figure rather than a recommendation. The model needed a source for the fact, not a brand.
- The prompt is category-level. When a user asks which tool to choose, the model tends to name brands that many independent sources associate with the category. Off-site presence in third-party pages therefore influences whether you are named, because the answer reflects outside corroboration, not what your own page says about you.
- The brand is a weak or ambiguous entity. A generic name, a name shared with other companies or inconsistent spelling across the web makes it harder for the model to treat your name as a distinct answer.
How ChatGPT, Perplexity and Google AI features show sources versus names
Engines differ in how they attach and display sources and in what their vendors document about inclusion. A gap measured in one engine therefore does not transfer to another. Report each engine separately. The table reflects documentation checked on 24 October 2026.
| Engine | How sources are shown | Documented inclusion requirement | What the vendor says about naming brands | What this means for the unnamed-citation gap |
|---|---|---|---|---|
| ChatGPT search | Linked sources on answers that use web search; help center wording not verified | Pages opted out of OAI-SearchBot are not shown in search answers, but can still appear as navigational links | Nothing documented | Confirm crawler access first; answers from trained knowledge can name you with no link |
| Perplexity | Linked sources alongside answers; help center description not verified | Not verified for this article | Not verified | Measure it on its own; ChatGPT results do not carry over |
| Google AI Overviews and AI Mode | Supporting links shown with the AI response | Page must be indexed and eligible to be shown in Google Search with a snippet | Nothing documented | A page excluded from snippets cannot be a supporting link, named or not |
Google's page on AI features, last updated on 10 December 2025 and consulted on 24 October 2026, states: "To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet".
We also infer that an AI Overview built from several pages can cite yours for one sub-point while the brands it names come from other pages. In that case the unnamed citation says little about how Google sees your brand.
What an unnamed citation is still worth: clicks, referrals and the metrics to track
Do not assume an unnamed citation brings visitors. A Pew Research Center analysis published on 22 July 2025 found that "Google users who encountered an AI summary also rarely clicked on a link in the summary itself. This occurred in just 1% of all visits to pages with such a summary." That figure concerns Google's AI summaries, not ChatGPT, so treat it as an indication, not a ChatGPT benchmark.
Measure the traffic directly instead. In analytics, filter sessions whose source or referrer contains chatgpt.com and look at the landing pages: if those pages receive almost none, the citation mostly serves someone else's recommendation. We could not verify OpenAI's publisher documentation on referral tagging on 24 October 2026, so check in your own reports how these visits are labelled. The setup is covered in setting up an AI referral channel in Google Analytics.
For reporting, track mention rate and citation rate as separate metrics rather than one visibility number:
- Mention rate: the share of runs in which your brand name appears in the answer text.
- Citation rate: the share of runs in which a page from your domain appears as a source.
- Cited-but-unnamed share: among runs that cite you, the share that leave your name out.
- Referral sessions: visits from ChatGPT to the cited pages.
These signals also explain why an AI visibility score can change while citations stay flat. A score that blends names and links moves when naming changes. Hypothetical example: your pages are cited as often this month as last, but a competitor is now named where you used to be. The citation count is unchanged and the blended score falls.
Answers also vary from one run to the next and from one day to the next, so a single check is a sample of one. Track results by re-running the same prompts over time.
What to change on the page and off it so the name travels with the citation
Each lever below is tagged as documented, observed in a study or inferred. No published case evidence shows a measured change in naming rates from these changes, including adding the brand name to passages; we found none when we checked on 24 October 2026.
- Name the brand in the passages most likely to be lifted (inferred). Put the name explicitly in the opening part of the page and in the first sentence of each key section, and spell it the same way every time. Write "Acme Tracker logs billable hours", not "Our tool logs billable hours" (Acme Tracker is hypothetical).
- Write self-contained passages with an explicit subject (inferred). A paragraph should make sense when read alone. Replace "it" and "we" with the brand name where a paragraph opens, so a lifted passage still carries the brand.
- Build third-party presence where the category is discussed (inferred). Reviews, comparisons, industry roundups and practitioner forums are where independent corroboration accumulates. This addresses the category-level cause described above.
- Add Organization markup for entity clarity (documented for Google, not for ChatGPT). Google's Organization documentation, updated on 8 September 2026 and consulted on 24 October 2026, states that "Adding organization structured data to your home page can help Google better understand your organization's administrative details and disambiguate your organization in search results." Include your official
nameandsameAslinks to your official profiles. Treat it as a clarity aid: it is not documented as a ranking or naming factor for ChatGPT. - Treat content optimisation results as a ceiling (observed in a study). In the GEO paper's abstract, consulted on 24 October 2026, Aggarwal et al. write that "GEO can boost visibility by up to 40% in generative engine responses." That is a maximum observed in their benchmark, not an expected gain, and it says nothing specific about brand names.
Adding the brand name to passages is a hypothesis to test, not a proven fix. The audit in the next section is how to test it on one page before rolling it out.
Where to start this week: audit which pages ChatGPT cites without your name
- Check crawler access first. Confirm that robots.txt does not block OAI-SearchBot on the pages you care about. Diagnose nothing else until this is clear.
- Pick and run prompts. Choose a small set of non-branded category prompts your buyers would type. Run each several times with search on and memory off, over several days, before treating any rate as real. Log the date, model, prompt, cited URLs and whether the brand is named, cited, both or neither.
- Verify every unnamed citation. Open the cited page and check that it contains and supports the claim the answer attributes to it. A citation can be wrong, and a misattributed page needs a different fix.
- Compute the cited-but-unnamed share per page and check
chatgpt.comsessions landing on those pages. - Rewrite one page and re-test. Take the page with the highest share, name the brand explicitly in its key passages, change nothing else and note the date. After it has been recrawled, re-run the same prompts with the same settings and compare.
- Report misrepresentation. If an answer misrepresents a cited page, use the feedback controls on the response. On 24 October 2026 we found no OpenAI documentation of a process for requesting attribution.
To see which of your pages ChatGPT, Perplexity, Gemini and Claude cite and whether your brand is named alongside them, run a free Namedrop scan as your next step.
Sources
- OpenAI, Overview of OpenAI Crawlers, consulted 2026-10-24
- Google Search Central, AI Features and Your Website, consulted 2026-10-24
- Google Search Central, Organization Schema Markup, consulted 2026-10-24
- Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, consulted 2026-10-24
- arXiv (Aggarwal et al.), GEO: Generative Engine Optimization, consulted 2026-10-24