You may repeat this evaluation your self fairly simply.
Simply do an open database search in Model Radar, head to the “Cited pages” report in your desired AI assistant, and export the highest 1,000 cited pages.
Then run the cited pages by means of our Batch Evaluation instrument to seize extra natural knowledge on every URL.
You need to use Claude to assist analyze that knowledge, and even write scripts for Google Colab to fetch and parse content material freshness indicators for every cited URL.
I did precisely that. Right here’s what I discovered…
ChatGPT’s citations aren’t confined to encyclopedic or editorial pages—however Wikipedia stands out as the one most-cited content material kind, far forward of all others.
The remainder of the citations are unfold throughout instructional content material, homepages, app listings, blogs, and different codecs.
I used Claude to categorize this knowledge. It is probably not 100% fool-proof, however it offers a directional understanding of ChatGPT’s most-cited pages.
Right here’s what the highest 1,000 cited URLs seem like damaged down by content material kind:
Wikipedia performs a central position in ChatGPT’s quotation habits. The assistant is clearly searching for structured, reference-style sources that summarize subjects comprehensively and predictably.
Academic and homepage pages additionally crop up frequently.
However what’s most attention-grabbing is the dearth of alternative to truly get talked about in these high-visibility citations.
Our Director of Content material Advertising and marketing, Ryan Regulation, refers to Wikipedia, homepages, app retailer pages as “lifeless” citations, in that you could’t simply affect them.
Going by the above categorization, solely 32.3% of the highest 1,000 citations in ChatGPT are pitch/outreach-worthy—or at the very least off-site and influenceable.
In different phrases, roughly two-thirds of ChatGPT’s prime citations are successfully off-limits to conventional outreach ways—they’re organizational pages, reference websites, and different “lifeless” citations you may’t realistically affect.
The pages ChatGPT cites are sometimes newer and fewer established in search, suggesting recency influences quotation visibility, although it’s not the one issue.
Amongst pages with detectable dates (~40% of the entire) the median age was 55 days (just below two months).
Listed below are the most-cited pages by publication yr…
And right here’s the breakdown of pages with detectable replace dates…
| Up to date | % |
|---|---|
| 30 days | 42.7% |
| 1-6 months | 20.0% |
| 6-12 months | 11.9% |
| >1 yr | 25.4% |
The info signifies a transparent recency bias, per findings from Metehan Yeşilyurt’s latest analysis.
Yeşilyurt recognized a URL_freshness_score in ChatGPT that favors newer content material, and cited research displaying that artificially refreshing publication dates can enhance AI rating positions by as a lot as 95 locations.
That stated, my knowledge confirmed almost one in 4 cited pages have been up to date over a yr in the past—suggesting that helpful, well-structured older content material can nonetheless preserve visibility in AI.
Practically one-third of ChatGPT’s citations level to pages with no conventional search visibility.
Listed below are some doable explanations for this:
Freshness: Some zero-keyword pages are doubtlessly recent content material not but ranked by serps. ChatGPT could uncover and cite new content material earlier than it accumulates search rankings.
Area of interest subjects: Citations could cowl particular, long-tail subjects with minimal search demand. They reply questions precisely however don’t appeal to vital search site visitors. This additionally applies to fan-out queries—even when somebody asks ChatGPT a few well-liked subject, the dialog typically branches into extra particular subtopics the place few pages straight goal these angles.
Completely different discovery: ChatGPT tends to entry and consider content material in another way than serps, seemingly prioritizing accuracy, freshness, and relevance over reputation indicators like backlinks.
The fact is it’s seemingly a mixture of all three. Recent content material, area of interest subjects, and various high quality indicators all contribute.
The pages that do rank show some attention-grabbing patterns. Listed below are the important thing takeaways:
- Web site authority actually issues: 65.3% are DR 81+, median DR 90
- However web page authority doesn’t: 67.3% have UR 0-10 and a median UR solely 6. ChatGPT cites pages from authoritative domains, however not essentially probably the most linked-to pages on these domains.
- Pages are keyword-rich: Median 279 key phrases
- Most have sturdy backlink profiles: Median 70 referring domains
- Half have excessive search visibility: 52.1% rank in prime 3 for key phrases
- Low-authority exceptions exist: 11.7% have DR 0-20
ChatGPT closely favors pages from high-authority domains which have sturdy backlink profiles and rank properly in search, however will cite low-authority sources for particular related content material.
Wrapping up
ChatGPT’s prime citations skew towards newer content material, embrace nearly a 3rd of pages with no natural visibility, and are dominated by reference websites you may’t simply affect.
The info suggests your finest wager for ChatGPT visibility is getting talked about in recent, particular content material that falls into these influenceable classes.
To search out outreach alternatives, you need to use Model Radar. Simply drop in your market or area of interest, head to the cited pages report and look out for the highest blogs, publications, and evaluation websites.
