You can rank #1 on Google and be invisible to AI

Most ChatGPT citations don't rank in Google at all

Ahrefs analysed ChatGPT's top 1,000 most-cited pages via Brand Radar in September and October 2025. visibility - not buried deep, not page two, absent from the SERPs. Zooming out to the query level: only 6.82%'s top 10 for the equivalent search query, and only 16.61%'s organic results. That means 83.39%'s results for the same query at all.

Domain authority gets the citation; page authority barely matters

The Ahrefs data reveals a sharp split in what predicts AI citation. 65.3% of ChatGPT-cited pages come from domains with a Domain Rating of 81 or higher - the median cited domain has a DR of 90. These are institutional reference sites, major publications, and established platforms. But here is the counterintuitive part: page-level authority barely matters. 67.3% of cited pages have a URL Rating of 0–10. A high-authority domain hosting a low-authority page still gets cited. Your brand-new long-form guide on a DR90 site can outperform a well-linked page on a mid-market domain.

There is also a structural bias in who gets cited. Ahrefs found that 67% of ChatGPT's top 1,000 citations are "off-limits" to most marketers - organisational reference pages, Wikipedia-style entries, government sources, established editorial outlets. Brands are 6.5× more likely to be cited via third-party coverage than via their own domain. A product page rarely gets cited; a product review on a trusted outlet does.

Position 5 gains more from optimisation than position 1

" - tested nine content tactics across 10,000 queries on a purpose-built benchmark. Adding statistics and citing authoritative sources lifted AI citation rates by roughly 41% on average. But the gains were not evenly distributed. Pages ranked around position 5 gained +115% lift. Pages at position 1 barely moved. This is almost the inverse of traditional SEO, the marginal return on content signals is highest for pages that are not already dominant.

Not all AI products work the same way

rank and AI citation is most pronounced in standalone, found citations come from pages in the top 10 or AI Mode at all.

citations and top-10 rankings at just 17–38%'s own included.

What the data does not say

The divergence is real, but three things should temper any headline reading of it.

fundamentals still gate AI inclusion.'s AI features require snippet-eligibility. "Not ranking organically" and "invisible to all AI" are not the same thing - many zero-SERP pages cited by ChatGPT are heavily linked reference pages that simply target informational queries that don't map cleanly to commercial keyword searches.

Some zero-visibility citations are training-era artifacts. ChatGPT's training data has a cutoff, and some high-citation pages reflect that snapshot rather than any repeatable optimisation signal. You cannot reliably reverse-engineer these citations into a strategy.

Earned media bias runs deeper than any single tactic..

Two separate jobs, tracked separately

rank and AI citation are now distinct metrics. A brand that only watches its SERP position is flying blind on roughly half its discoverability - most brands have no idea what they are missing. You need to know whether the models mention your brand at all, which sources they draw on when they do, and whether the sentiment they carry is the message you want amplified. Baseline Labs runs that scan across ChatGPT, Perplexity and Gemini - the same way we did for the stout brands paper measured citation rates. Run one on your brand to see where you actually stand.

Scan your brand's AI visibility