Only 7.8% of the URLs that ChatGPT cites overlap with Google's top organic results. For Claude, it's 11.2%. Even Google's own AI mode only matches its traditional rankings about a third of the time.
Those numbers come from a study by AI Plus Automation that ran 120 queries across four major AI platforms - ChatGPT, Claude, Perplexity, and Gemini - then crawled 479 cited pages to figure out what actually gets referenced. The results should make anyone rethinking their content strategy for AI visibility pay close attention.
The Factors That Actually Predict Citations
The study found seven statistically significant predictors of whether a page gets cited by AI systems. The strongest? Internal links, with an odds ratio of 2.07 - meaning pages with robust internal linking structures were roughly twice as likely to be cited.
The other signals that mattered: word count (cited pages had a median of 2,582 words), schema markup (the structured data that helps machines parse your content), self-referencing canonical tags (present on 84.2% of cited pages), and content-to-HTML ratio.
Here's what didn't matter, despite years of SEO advice saying otherwise: author bios, page speed, popup presence, and paywalls. None showed statistical significance. Author attribution was actually slightly lower on cited pages (44.0%) compared to non-cited ones (48.7%). So much for E-E-A-T being the golden ticket to AI visibility.
How Each Platform Handles Citations
The four platforms behave very differently under the hood, and that matters for anyone trying to get cited.
Perplexity cites sources on 97% of responses, averaging 9.8 sources per query. Google's AI mode is even more citation-heavy at 98%, pulling in 17.4 sources per answer. ChatGPT sits at 56% with 3.5 sources, while Claude is the most selective at 39% with 5.5 sources per query.
There's also a fundamental architectural split. ChatGPT and Claude are "live-fetch" platforms - they send real HTTP requests to websites during conversations. Perplexity and Gemini use pre-built indexes and never contact your server while generating answers. This means your server logs might show visits from ChatGPT's crawler, but Perplexity's citations come from a snapshot it already has.
What This Means For Your Content
The biggest takeaway isn't about any single technical factor. It's that optimizing for Google and optimizing for AI citations are increasingly separate games. When ChatGPT overlaps with Google's results less than 8% of the time, ranking #1 on Google doesn't mean ChatGPT will ever mention you.
The study found that intent matching matters more than page-level optimization. Your content needs to align with what each AI system expects for a given query type - discovery, informational, or product comparison - before any technical tweaks move the needle.
For practical purposes: long-form content with strong internal linking, clean schema markup, and proper canonical tags gives you the best shot across all four platforms. But the days of a single SEO strategy covering both traditional search and AI answers are over. These are different systems pulling from different pools of content, and treating them as one channel is a mistake.