The sources AIs cite most when talking about brands (123 audits)

When an AI talks about your brand, where does it get its information? We looked, on real data, at which sources models cite most often. The answer changes how you work on your visibility.

The scope

We analysed the sources cited across 123 audits (7,264 AI answers): the addresses models present as sources for their answers about brands. Each source is counted once per audit (not by raw citation count), to measure cross-cutting authority rather than a single brand's weight. Brands' own websites are excluded: we keep only third-party sources.

The ranking — where AIs look

Here are the most frequently cited third-party sources, as a share of the 123 audits in which they appear.

Finding 1 — Your website is not enough

This is the core message: AIs do not build your brand's image from your website alone. They rely heavily on third-party platforms. LinkedIn comes first (cited in more than one audit in three), followed by Wikipedia and Trustpilot. If these surfaces say something inaccurate or incomplete about you, the AI repeats it.

Finding 2 — Business registries weigh heavily (especially in France)

A whole block of sources comes from business registries and directories: Pappers (24%), Societe.com (19%), PagesJaunes, the official data.gouv.fr registry, Kompass, entreprises.lefigaro. AIs draw the "hard" facts from them — legal name, directors, activity, seniority. A wrong or outdated legal detail in these databases ends up in the answers.

Finding 3 — Reviews and community feed the AI

Review platforms (Trustpilot, G2, Glassdoor, Clutch) and community spaces (Reddit, YouTube) come up constantly. In other words: what customers, employees and users say about you elsewhere becomes a direct source for the AI.

What it changes in practice

Working on AI visibility is not only about optimising your website. It is about controlling what the sources AIs cite say about you: an up-to-date LinkedIn page, an accurate Wikipedia presence when warranted, managed reviews (Trustpilot, G2), and correct company data in registries. That is precisely what our co-citation and source authority analyses measure.

The limits of these figures

This data comes from brands that chose to audit themselves — a sample not representative of all companies, with a marked French component (hence the weight of FR registries). The ranking reflects citation frequency, not a source's "quality". See also our barometer, our study on invented URLs, the Observatory and the methodology.

Want to know which sources AIs cite for your brand — and which ones hurt you? That is exactly what an audit measures.