Skip to main content
AI Business, Funding & Market

Preferred Sources

Definition

A Google feature that surfaces and labels sites a reader has personally designated, both in Search results and in AI answers. It is a rare point where a site owner can directly request AI visibility, though the designation itself belongs to the reader, which keeps it earned rather than owned

#Preferred Sources#AI Overviews#AI Mode#Google Search#AI visibility#GEO

What are Preferred Sources?

Preferred Sources is a Google feature that lets a reader designate the sites they want to see more of, after which Google surfaces those sites more prominently — and labels them — in that reader's search results and AI answers. It began in Top Stories in 2025, expanded to all supported languages on 2026-04-30, and reached AI Overviews and AI Mode on 2026-05-27 (Google). As of that announcement, more than 345,000 unique sites had been selected. Site owners can install a button on their own pages that requests the designation.

Why does the distinction matter?

AI answer visibility is largely outside a site owner's control — nobody can tell a model what to search for or which document to cite. Preferred Sources is the exception: a channel for directly requesting visibility.

It should still be classified as earned, not owned. Installing the button does nothing until a reader chooses to press it, and the designation can be removed at any time. What the owner controls is the visibility of the request. That makes it structurally different from source gap work — a source gap is about entering third-party surfaces, while Preferred Sources reinforces reach among an audience you already hold.

Eligibility and setup

  • Domain or subdomain onlyexample.com and blog.example.com qualify; a subdirectory such as example.com/blog does not. You cannot nominate one section of a site.
  • No industry restriction — the publisher documentation contains no news-publisher-only clause.
  • No structured data required — a JavaScript button or a deep link is sufficient.
  • Language-agnostic — available in every language where Google Search operates.

Measurement notes

  • Personalization makes it hard to isolate — readers who did not designate you see no change, so aggregate impression metrics will not separate the effect. Pair brand-search traffic with return-visit metrics instead.
  • Google's "twice as likely to click" is vendor-reported — sample and study design were not published, and readers who designate a site already trusted it, so the direction of causality is unestablished. Read it as correlation, not causation.
  • It is not a discovery channel — it acts on people who already know you. Reaching those who do not still depends on content selected as an answer's basis and on citations in third-party documents.
  • Placement drives conversion — first-time visitors do not press the button. Newsletter confirmation screens and the end of a content series convert better than a site-wide banner.

To see how AI answers currently treat your brand, run an AI brand visibility analysis; to check the site side, use the page structure diagnosis.

Is your site visible in AI search?

See for free how ChatGPT, Perplexity, and Gemini describe your brand.

Start Free Diagnosis →

Related terms