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. 28.3% had zero organic Google visibility — not buried deep, not page two, genuinely absent from the SERPs. Zooming out to the query level: only 6.82% of pages ChatGPT returns appear in Google's top 10 for the equivalent search query, and only 16.61% appear anywhere in Google's organic results. That means 83.39% of ChatGPT's retrieved pages don't appear in Google's results for the same query at all.

Authority at the domain level, not the page level

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

The Princeton / Georgia Tech GEO paper — published at KDD 2024, coined the term "Generative Engine Optimization" — 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, where top-ranked pages compound their advantage. In GEO, the marginal return on content signals is highest for pages that are not already dominant.

Not all AI products work the same way

The divergence between Google rank and AI citation is most pronounced in standalone LLMs like ChatGPT. Google's own AI features behave differently — at least for now. An Ahrefs study by Ryan Law in July 2025, covering 1.9 million citations across 1 million AI Overviews, found 76% of AI Overview citations come from pages in the top 10, with cited URLs sitting at a median SERP position of 3. Google has explicitly stated its AI features are "rooted in our core Search ranking and quality systems," and a page must be indexed and snippet-eligible to appear in AI Overviews or AI Mode at all.

That said, even this correlation is weakening. By February 2026, SEJ and Ahrefs data put the overlap between AI Overview citations and top-10 rankings at just 17–38% — down sharply from the July 2025 figures. The direction of travel is clear: AI citation is detaching from traditional rank across every major product, Google's own included.

What the data does not say

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

SEO fundamentals still gate AI inclusion. A page must be indexed and crawlable before any AI product can cite it. Google'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. Chen et al. (arXiv:2509.08919, Sept 2025) found AI search shows "a systematic and overwhelming bias towards earned media — third-party, authoritative sources — over brand-owned and social content." The implication: being covered by trusted outlets matters more than optimising your own pages. That is a different kind of problem than traditional on-page SEO.

Two separate jobs, tracked separately

Google rank and AI citation are now distinct metrics. A brand that only watches its SERP position is flying blind on roughly half its discoverability. 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, the same way the GEO paper measured citation rates. Run one on your brand to see where you actually stand.

Scan your brand's AI visibility