Do all AI engines cite the same sources?
No. ChatGPT, Gemini, Copilot, and Perplexity each pull citations from measurably different pools, and most of what they cite does not rank in Google's top ten. Treating the engines as one audience is the most common mistake in generative engine optimization: each engine runs its own retrieval, trusts its own sources, and rewards different signals.
The numbers below come from two 2026 measurement sets: Ahrefs' overlap studies of citation behavior across assistants and AI Overviews, and Muck Rack's analysis of which publications each model actually reads. Together they replace guesswork with a map.
Ranking on Google no longer buys the citation
Only 38% of pages cited in Google AI Overviews also rank in the traditional top ten, down from 76% seven months earlier (Ahrefs, 863,000 keywords and 4 million cited URLs). The gateway theory of AI visibility, rank first and the citation follows, is collapsing in the data.
Across AI assistants the link is weaker still. When Ahrefs asked ChatGPT, Gemini, Copilot, and Perplexity 15,000 real questions, only about 12% of the URLs they cited ranked in Google's top ten for the matching query (Ahrefs). More than 80% of citations came from pages that do not rank. Rankings remain worth winning, but they are one door into the answer, not the door.
Perplexity is the Google loyalist
Perplexity cites Google's top ten almost four times more than the other assistants: 28.6% of its cited URLs rank there, against roughly 8% for ChatGPT, Gemini, and Copilot (Ahrefs).
The implication is direct. If your pages already rank, Perplexity is where that equity pays out fastest. On the other engines, rankings barely register: visibility there is earned with third-party mentions and pages structured for extraction, not positions.
Each engine keeps its own reading list
Claude cites Reuters fifty times less often than ChatGPT does, and twenty times less often than Gemini (Muck Rack, What is AI Reading?). Its most-cited list favors evergreen, analytical sources: Harvard Business Review, TechRadar, Good Housekeeping. Freshness splits the same way: 56% of ChatGPT's journalism citations were published within the past year, against 36% for Claude.
The practical reading: a news-cycle PR win feeds ChatGPT and Gemini. A definitive evergreen explainer feeds Claude for years. A citation strategy built for one engine's diet leaves the others hungry.
What this means for an exporter in Iraq or the Gulf
Engine-segmented visibility work beats a single citation strategy: the engines your buyers use decide which sources must name you. A distributor in Dubai asking Perplexity rewards the Google rankings you already hold. A procurement lead pasting a shortlist into Claude rewards the evergreen explainer that names your company. A buyer on ChatGPT rewards recent third-party mentions.
This is the program we run inside our generative engine optimization services: extractable pages on your own domain first, then mentions on the sources each engine actually reads, tracked per engine. For companies selling out of Iraq and the Kurdistan Region, that work starts with SEO in Iraq for companies that sell abroad.
