How to Rank in ChatGPT: What It Actually Cites (I Tested It)
Summary
Matthew Zare ran a hands-on test of ChatGPT: he asked it real content marketing questions with search turned on and opened every page it cited. One answer placed an established Ahrefs article beside a short page that had no author and no date. With search off, ChatGPT cited nothing. Zare uses the result to explain how the assistant picks sources, drawing on Ahrefs research on freshness and brand visibility, the Princeton GEO paper, and Google's spam policy.
He attributes the short page's inclusion to recency and to how easy it was to quote. In his sample, 4 of the 5 cited pages were dated 2026. For recommendation prompts, he found that only already well known brands were named. His advice is to show your method (who wrote it, what you tested, and when) and to publish original numbers that others will cite.
Demonstration by Matthew Zare, embedded from YouTube. AIOTruth did not reproduce the test.
Why it matters
Getting cited by ChatGPT depends on the assistant running a search, so a page that OAI-SearchBot cannot crawl is out of the running before content quality matters. Inside a search, freshness and quotable, direct answers can lift a small page next to a large publisher. Brand recommendations follow a different pattern: they track how widely a brand is already mentioned across the web and YouTube, not how many pages it publishes. Teams can track ChatGPT referral clicks through the utm_source=chatgpt.com tag. They should not chase citations with thin pages, because Google's spam policy now covers attempts to manipulate its AI answers.
Source
- developers.google.com/search/docs/essentials/spam-policies the item itself, published by Matthew Zare, created by Matthew Zare
- arxiv.org/abs/2311.09735 Original dataset with disclosed methods
- ahrefs.com/blog/do-ai-assistants-prefer-to-cite-fresh-content/ Reputable reporting
- ahrefs.com/blog/ai-brand-visibility-correlations/ Reputable reporting
- ahrefs.com/blog/branded-web-mentions-visibility-ai-search/ Reputable reporting
- www.orbitmedia.com/blog/blogging-statistics/ Reputable reporting
- youtube.com/watch?v=HgUqORcD7yAwhere AIOTruth found it
What was checked
- canonical URL taken from the page itself
- read the item page (200)
- https://arxiv.org/abs/2311.09735: supports the item, tier 5 (Original dataset with disclosed methods)
- https://ahrefs.com/blog/do-ai-assistants-prefer-to-cite-fresh-content/: supports the item, tier 9 (Reputable reporting)
- https://ahrefs.com/blog/ai-brand-visibility-correlations/: supports the item, tier 9 (Reputable reporting)
- https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search/: supports the item, tier 9 (Reputable reporting)
- https://developers.google.com/search/docs/essentials/spam-policies: supports the item, tier 1 (Official platform documentation)
- https://www.orbitmedia.com/blog/blogging-statistics/: supports the item, tier 9 (Reputable reporting)
Limitations
- The demonstration was not independently reproduced by AIOTruth.
AIOTruth judgment
The item earned a Value Score of 9.3 with an evidence status of Confirmed. The fetched sources back its main claims: the Ahrefs freshness and brand visibility studies, the arXiv GEO paper with disclosed methods, Google's own spam policy documentation, and the Orbit Media survey. Evidence and usefulness both scored 10, because every claim points to a checkable source and the advice can be applied directly. Originality scored 9 because the test is first-hand rather than a summary of other work. Relevance scored 8. The recorded limitation still applies: AIOTruth did not independently reproduce the demonstration, and the presenter calls the brand finding a pattern, not a law.
How this score was calculated
| Dimension | Weight | Score | What produced it |
|---|---|---|---|
| AIO relevance | 30% | 8 | ai search, generative engine optimization, geo, ai overviews, ai mode; 4 scope question(s) matched |
| Usefulness | 30% | 10 | 3 artifact(s), 6 actionable marker(s), 2 measured figure(s) |
| Evidence | 25% | 10 | Official platform documentation; 6 verified source(s) across 4 domain(s) |
| Originality | 15% | 9 | carries original testing or data; closest archive match 0.19 |
aioRelevance x 0.30 + usefulness x 0.30 + evidence x 0.25 + originality x 0.15. The rubric is published in full on the Editorial Method page. Scoring is deterministic: the same item scores the same on every run.
Watch the demonstration Open the source
Found something wrong here? AIOTruth corrects material errors openly. Have something we should review? Submit a find.