Skip to content

Observatory · Engine

Which sources does Claude cite on buying questions?

Claude takes 88.1% of its cited sources from rankings and comparison pages on buying questions, the highest share of the six engines in the Observatory. Data: 8,020 answers collected via Peec AI between July 12 and October 11, 2026.

Updated

84%

answers citing sources

4.2

sources per answer

88%

rankings among sources

8,020

answers analysed

Types of sources cited : Claude
Rankings and comparisons88%Other sites9%Reviews and directories3%Forums (Reddit, Quora)1%

Data : Peec AI, 8,020 answers analysed, July 12, 2026 to October 11, 2026.

Claude is the engine that relies most on third-party rankings. Across 8,020 answers to buying questions, mostly in French, 88.1% of the sources it cites are rankings, "best of", comparison or "alternatives" pages. Brand sites, media and blogs weigh only 8.7%, reviews and directories 2.6%, forums 0.6%. YouTube, LinkedIn and Wikipedia are at 0.0%.

The page types confirm it. Listicles make up 55.7% of the pages Claude cites, comparisons 11.4% and "alternatives to" pages 10.9%. Product pages (5.9%), category pages (4.4%) and homepages (3.8%) trail far behind, followed by profiles (2.3%), articles (1.8%) and how-to guides (1.0%).

Claude cites at least one source in 84.1% of answers, with 4.2 sources cited and 6.8 consulted per answer. On identical questions asked to both engines, it takes 87.8% of its sources from rankings against 57.0% for ChatGPT (ChatGPT vs Claude).

A limit to keep in mind: the Claude sample (8,020 answers) is smaller than the Profound engines, and it has no breakdown by sector or month. Its direction is clear, though: to appear in Claude, a brand must first appear in the lists Claude reads. That is the core of our Claude optimization service and of our authority and citations work.

Method

406,626 AI answers collected every day from July 12, 2026 to October 11, 2026 on over 1,700 purchase-intent questions tracked by Archipel (mostly in French, plus the United States and the Netherlands). ChatGPT, Google AI Overviews, Perplexity and Copilot: Profound data. Claude and Gemini: Peec AI data. Sources are classified automatically from their address and title. Figures are aggregated: no brand or client is published. Free to reuse with a link to this page.

FAQ

Frequently asked questions

Which sources does Claude cite on buying questions?
On buying questions, Claude cites rankings and comparison pages for 88.1% of its sources, brand sites, media and blogs for 8.7%, and reviews and directories for 2.6%, based on 8,020 answers tracked via Peec AI.
Does Claude cite listicles and "best of" articles?
Yes, Claude cites listicles more than any other page type: 55.7% of the pages it cites are listicles, plus 11.4% comparisons and 10.9% "alternatives" pages.
How many sources does Claude cite per answer?
Claude cites 4.2 sources per answer on buying questions and consults 6.8. It cites at least one source in 84.1% of its answers.
Does Claude rely on rankings more than ChatGPT?
Yes, on identical questions asked to both engines, Claude takes 87.8% of its sources from rankings against 57.0% for ChatGPT, based on 5,174 answers per engine.

Shall we get started?

Free GEO audit: your rank on ChatGPT, Claude and Google AI Overviews, your competitors and your first actions. Reply within half a business day.