Google Preferred Sources: personal choice, not a universal ranking promise
Preferred Sources lets readers choose sites for more prominent relevant coverage in Top Stories. Google reported over 200,000 selected sites and twice the click-through likelihood on 30 April 2026; those are dated aggregate observations, not a forecast for your business. The same date brought all supported Search languages. Current guidance also covers eligible AI experiences. This article separates those documented functions from unmeasured claims.
~/.claude/skills/ or into .claude/skills/ within your project. Start Claude Code with claude, then enter /brand-mentions-inventory inside the session. Automatic discovery and model compliance are not guaranteed.AGENTS.md, or add it to your repo as a separate file and refer to it from AGENTS.md. Codex reads that with every session.
How a preferred source works
reg. P.001Preferred Sources records a reader’s choice of sources; it is not a subscription or a public endorsement score. The four steps below describe the documented Top Stories route. Content must still be relevant, and a personal preference is not evidence of a ranking boost for every other searcher.
From experiment to every language
reg. P.002Google documented a Labs experiment in June 2025, the US/India launch in August 2025, the worldwide English-language expansion in December 2025 and all supported Search languages on 30 April 2026. That is less than a year from the first listed stage to the language expansion, not more than a year. This timeline records announcements, not adoption rates or a business result.
Do you qualify?
reg. P.003The source selector accepts domains and subdomains, not a subfolder as an independently selectable source. A blog in a subfolder can still belong to its selectable parent domain. Check the actual source preferences tool; a domain-shaped URL alone does not establish availability. AI display also requires inclusion in Search generative AI features through Search Console under the current guidance.
Where is your brand already mentioned?
reg. P.004Familiarity may encourage a reader to choose a source, but mentions do not automatically produce that choice. The download inventories owned, third-party and machine discovery channels before proposing work. Each finding needs evidence; not checked and not found within a specified search are different statuses. AI can help organise the analysis without inventing the observations.
What a source preference does and does not measure
Technology, content and reputation remain useful areas of SEO work. Preferred Sources adds an explicit reader choice, not a measured universal trust score. Google’s E-E-A-T guidance helps evaluate content, but human raters do not directly set rankings. A preference records a choice in a specific feature; it does not reveal every reason for that choice or establish that this is the first measurable expression of trust.
Reputation research is useful, with a separate evidence boundary
In those same rater guidelines, Google asks its raters to look beyond a site at what others write about a company. Human reputation research and a user-selected preference have different functions. Google says rater data is not used directly in ranking algorithms. And with the rise of AI Overviews and zero-click search, the question of who gets cited and trusted only becomes more important: an AI answer may cite sources but is not guaranteed to cite this brand, and the Preferred Sources documentation explicitly states that chosen sources can carry a preferred label in AI Overviews and AI Mode for the searcher who chose them.
Branded search is another signal to investigate, not proof of trust
Brand familiarity may relate to name searches, but search volume alone cannot establish trust or its cause; negative publicity and other factors can also prompt a search. How you become visible in AI answers, and why that rests on the same foundations as classic findability, is covered in more depth in getting found in AI Overviews and ChatGPT NL. A sampled view of recorded model answers is offered by the AI Index NL: its method uses sector questions and counts mentions in returned answers. Record model, query, date, locale, search mode, context and trial count; no test of this reader’s business has been run here. Want that measured and improved for your own brand? That is AI visibility.
Important to add: none of this makes Preferred Sources a classic ranking factor
The documented feature personalises relevant appearances for the person who selects the source. It does not establish an ordinary-ranking boost for everyone. Familiarity, selection, repeat reading and business value are separate stages that need their own evidence; none automatically follows from the previous stage, and returning readers are not universally the most valuable visitors.
What you can already do today
First check whether the site appears in the source selector. Google currently recommends an automatically localised JavaScript button and also documents a deep link: google.com/preferences/source?q=yourdomain.com. Place an appropriate invitation where readers can understand it, such as an article footer. The deep link is an alternative when the script cannot be used, not an automatic subscription. Check the current supported-language list and explain that the reader chooses and confirms.
The second place is your newsletter
Subscribers have chosen an email relationship, not automatically a Search preference. Offer an optional, explained invitation without assuming uptake. Put the link in your welcome email and repeat it occasionally with a short explanation. The same applies to any contact moment you already have: if you ask customers for a review once a job is finished, for example as described in asking for a Google review via WhatsApp NL, a separate preference invitation might fit an appropriate contact. Do not silently add it to the download’s one-question review-request template or assume contact permission; decide the message’s purpose and scope first.
Also take care of the places where that trust begins
Your Google Business Profile is often a local business's first introduction to your brand; a complete profile with real photos does quiet work there. And the site itself has to deliver what a preferred source promises: publishing something worthwhile on a regular basis. Watch out here for the pitfall of the big promise: a fixed publishing rhythm is valuable, but do not promise yourself a weekly frequency you cannot keep up. Three months without publishing is not a documented universal exclusion threshold. Relevance and freshness matter for Top Stories, but older material is not thereby proven invisible across all Search.
Finally, the basics
Preferred Sources does not compensate for technical problems: your site has to be indexed and simply work well. If you are unsure about that, first have the free SEO scan look at your site or check the fundamentals via technical SEO. How all of this fits into a complete approach, from technology to brand building, is covered under improving your SEO.
Mentions: the family backlinks belong to
Anyone who has been in SEO for a while recognises the pattern: becoming a preferred source looks suspiciously similar to earning a good backlink. In both cases, someone else expresses trust in you, in both cases you cannot force it, but neither analogy establishes identical ranking mechanics or enforcement. The difference is who chooses: with a link, an editor or site owner; with a preferred source, the reader themselves. What a backlink actually is, and why it still counts, you can read in what is link building NL; the complete approach to earning more of them is covered in getting more backlinks.
The mention without a link has long been underestimated
A trade magazine article that names your company without linking, a guide that recommends you, a partner who mentions you on their site: they are all traces of the same trust, and human raters may examine external reputation, but this does not establish that every AI system uses each mention or that it is a direct ranking signal. That is why the skill accompanying this article starts not with links but with mentions in the broad sense. The inventory identifies evidence-backed possibilities for further review, not guaranteed placements: we take that work on within link building, and how an editorial placement comes about with us is covered under the backlink network NL.
And sometimes the outcome of the inventory is that you should not start with mentions at all
A site that answers the search query better than the competition can also climb without any new external signals; how that works is covered in ranking higher in Google without backlinks. The honest order stays the same as ever: first a site that is worth it, only then tell the world about it. Want more of this kind of well-founded explanation? Our knowledge base is full of it.
What we do not know for certain, and what is doing the rounds
Three claims need qualification. First, the cited Preferred Sources guidance covers Top Stories and eligible AI experiences; it does not establish the earlier claim about Discover. That is a limit of this evidence, not proof that no future feature can exist. Second, the guidance does not establish higher ordinary rankings for everyone. Third, worldwide totals are not country-specific counts. The over-200,000 figure belongs to Google’s April announcement, not an audited current national total.
Google’s current Search Central guide documents source selection, domain/subdomain scope, interactive buttons, deep links and eligible AI display. Its 30 April 2026 announcement reports over 200,000 selected sites and twice the click-through likelihood after selection. These are Google-reported aggregate observations, not an independent causal experiment or a business forecast; they do not measure how likely your site is to be chosen.
This review does not establish: how heavily this signal will ever weigh outside personalisation, whether Google extends the feature to other parts of Search, and how quickly searchers will start using it. So be wary of anyone selling certainty here; how to spot such sellers is covered in choosing an SEO agency NL. For anyone who also wants to explain their site to machines, llms.txt is a related topic in the same category: promising, but tell the honest story alongside it.
Frequently asked questions
What are Google Preferred Sources?
Preferred Sources lets readers select sites to see more relevant content from them in Top Stories. The usual route is the star beside Top Stories and a source-selection panel. Google recorded all supported Search languages on 30 April 2026. Current guidance also allows a preferred label in eligible AI Mode and AI Overviews experiences; availability and site eligibility still need checking.
Does Preferred Sources also work in ordinary search results?
The cited guidance does not establish a universal boost to ordinary organic positions. It describes more prominent relevant Top Stories content and a possible preferred label in eligible AI Mode and AI Overviews for readers who selected the source. Do not infer an undocumented causal ranking gain from those functions.
Does my website qualify as a preferred source?
Check whether the domain or subdomain appears in Google’s source preferences tool. A subfolder such as yourbusiness.co.uk/blog/ cannot be selected separately, although its parent domain may be selectable. Indexed status and a domain-shaped URL alone do not confirm availability. Current AI-feature guidance also requires inclusion in Search generative AI features through Search Console.
How do I ask visitors to mark my site?
Google documents an automatically localised interactive JavaScript button, custom-design integration and a deep-link alternative at google.com/preferences/source?q=yourdomain.com. Check current language support rather than relying on the earlier sixteen-language asset count. Explain the choice clearly and let the reader confirm it; a link click alone is not proof of a saved preference.
Can I see how many people have chosen me as a preferred source?
The cited publisher guide does not provide a site-owner count of readers who selected the source. That is not proof that no measurement can ever exist. Search Console click-through changes alone cannot identify this feature’s effect because query mix, devices and other factors may change. Button clicks can measure interaction where appropriately implemented, but do not by themselves establish a saved preference.
Is this only interesting for news sites?
Top Stories makes fresh, relevant publications an obvious use case, but current guidance also includes eligible AI experiences. Google’s April 2026 announcement reported over 200,000 selected sites, from local blogs to major news desks. That dated total does not measure the gain for a non-news business or prove a small effect. Assess relevance and actual results rather than assume them.
Start with the inventory, not the button
The temptation is to place a deep link today and wait. Check eligibility and choose an appropriate implementation; five minutes is only an estimate. Selection is voluntary and is not restricted to a proven pre-existing trust score. The real work is building that trust, and that starts with knowing where your brand stands today: which mentions exist, which are wrong, and where competitors were verified and your brand was not found within the same dated search scope. The download’s four-hour example is a proposed plan with unknowns, not a live audit. Download the free skill at the top of this article for that, and let Claude or Codex build the inventory with you. Want us to look at the bigger picture of your findability with you? That is possible via the contact page or with a free scan; you will find all the other tools we have built under skills.