Google Preferred Sources for AI Overviews: What Publishers and Brands Should Know
Google’s Preferred Sources feature gives users a direct way to select websites they want to see more often in Search. Here is what the feature changes, what it does not guarantee, and how publishers should measure it.

What Google Preferred Sources changed
On August 20, Google announced an embeddable “Add to Preferred Sources” button. When a user selects a qualifying website, Google can use that preference to show the source more prominently for that user in relevant Search experiences.
Google says users have already selected more than 600,000 unique sources. That wording matters: it refers to sources selected by users, not necessarily 600,000 participating publishers or websites that have installed the button.
The feature is globally available, but a domain must appear in Google’s source-preferences tool to qualify. Publishers should therefore verify eligibility before planning implementation.
See Google’s announcement and implementation documentation for the current requirements.
How Preferred Sources works
Preferred Sources appears to be an explicit user preference. A reader chooses a source, and Google can use that choice to personalise future visibility. This is different from trying to improve a page’s general relevance, authority, or technical performance.
The practical distinction is simple:
- general ranking signals help determine what may be relevant broadly;
- Preferred Sources can influence what a particular user wants to see more often;
- AI citations depend on the answer system’s source selection for a specific prompt.
A publisher can make the preference available and encourage genuine readers to select the source. It cannot force selection, control every answer, or guarantee a citation.
Preferred Sources vs rankings vs AI citations
| Signal | Traditional ranking | Preferred Sources | AI citation |
|---|---|---|---|
| Main driver | Relevance, quality, authority, and context | Explicit user preference | AI system’s source selection |
| Personalised? | Sometimes | Yes | Sometimes |
| Controlled by publisher? | Indirectly | Only through eligibility, implementation, and audience adoption | No direct control |
| Guarantees visibility? | No | No | No |
| Useful measurement | Rankings, impressions, clicks, traffic | Selections, interactions, and personalised visibility | Mentions, citations, accuracy, and source inclusion |
Preferred Sources should therefore be measured separately from ordinary rankings and from AI citation rate. A change in one metric should not automatically be described as the cause of a change in another.
Why this matters for AI search visibility
It adds a user-controlled visibility layer
AI search visibility is usually discussed in terms of content, entities, authority, sources, and answer relevance. Preferred Sources adds a documented mechanism in which the user can express a direct source preference.
That does not replace content quality. It adds another condition worth measuring: whether a publisher has a real audience willing to select and return to its content.
It strengthens the value of a returning audience
Publishers may get more value from newsletters, LinkedIn audiences, research subscribers, and other owned distribution channels. Those channels can help readers discover useful content and, where appropriate, choose the publisher as a preferred source.
This makes audience development part of the broader AI visibility conversation. GEO is not only a page-level optimisation exercise.
It does not replace entity clarity or evidence
A preferred-source selection does not fix an unclear publisher identity, weak topical coverage, unsupported claims, poor internal linking, or inconsistent information across the web. Ranknizer’s working model still applies:
- diagnose how the source appears and where it is missing;
- strengthen clarity, coverage, evidence, and connected content;
- measure the same prompts and surfaces again.
How publishers should implement Preferred Sources
Use a staged process rather than adding the control everywhere immediately.
1. Check eligibility
Confirm that the domain appears in Google’s source-preferences tool. Do not assume that every website qualifies simply because the feature is globally available.
2. Add the official button to suitable content
If eligible, place the official button below articles, editorial guides, and research pages where a reader has already received value and may reasonably want more from the publication.
3. Promote the deeplink selectively
The deeplink may be suitable for a newsletter, LinkedIn profile, author profile, or selected social posts. The copy should explain the benefit without suggesting that selecting the source guarantees citations or rankings.
4. Protect conversion pages
Do not make the button prominent on service, pricing, audit, or tool-input pages if it distracts from the page’s primary action. A source-preference control should support audience retention, not compete with a purchase, enquiry, or diagnostic flow.
5. Establish a baseline before measuring impact
Record relevant Google AI visibility before implementation. Include the prompts, answer surfaces, competitors, mentions, citations, and accuracy observations that will be used for follow-up testing.
What Ranknizer should measure
Diagnose
Start with:
- domain eligibility;
- button impressions and interactions;
- the pages where the control appears;
- newsletter and social promotion;
- baseline visibility in AI Overviews and AI Mode;
- source mentions and citations;
- answer accuracy and commercial relevance;
- competitor visibility for comparable prompts.
Strengthen
Use the findings to improve:
- publisher and brand entity clarity;
- author and organisation information;
- useful, answer-ready content;
- internal links between related research and commercial pages;
- evidence and source quality;
- newsletter and social distribution;
- the placement and wording of the button.
Measure
Retest the same prompt set and separate the results by platform and outcome. Track preferred-source interactions independently from:
- organic rankings;
- AI Overview visibility;
- AI Mode visibility;
- mention rate;
- citation rate;
- answer accuracy;
- competitor share;
- commercial prompt visibility.
The purpose is not to claim that the button caused every visibility change. The purpose is to create a clearer evidence trail around a new audience preference signal.
What Preferred Sources does not mean
Preferred Sources does not guarantee citations. It does not improve rankings for users who have not selected the website. It is not automatically available to every domain, and the reported 600,000 unique sources should not be described as 600,000 enrolled publishers.
It also does not make FAQ schema a ranking or eligibility shortcut. Visible FAQs can help readers understand the topic, and appropriate structured data can describe the page, but neither should be presented as a guarantee of inclusion in AI answers.
A new audit opportunity for publishers and brands
Preferred Sources eligibility and implementation can become a focused item in an AI visibility audit. A useful review could include:
- eligibility verification;
- button and deeplink implementation;
- recommended placement by page type;
- audience-promotion opportunities;
- baseline and follow-up AI visibility testing;
- competitor and source-pattern comparison;
- conversion-risk review.
For Ranknizer, this fits naturally into the existing evidence-led approach: identify what changed, define the measurement boundary, and turn the finding into a prioritised action rather than a speculative ranking claim.
Frequently asked questions
Can users influence which websites Google AI Overviews cites?
Users can express a source preference, and Google may use that preference to show a selected source more prominently for them. The preference does not guarantee that the site will be cited in every AI Overview.
Does Preferred Sources improve Google rankings?
It should be treated as a personalised visibility feature, not a general ranking improvement. A selection may affect the selecting user’s experience, not the results of every searcher.
Is Preferred Sources available to every website?
No. A domain must appear in Google’s source-preferences tool to qualify. Publishers should check eligibility before implementing the button.
Should every page include the button?
No. It is most appropriate on useful editorial, news, and research content. It may distract users on service, pricing, audit, and tool-input pages.
Does adding the button guarantee AI citations?
No. Citations depend on the answer, prompt, platform, source selection, and other changing systems. The button creates an opportunity for user preference; it does not provide control over output.
Does FAQ schema affect Preferred Sources eligibility?
There is no basis for treating FAQ schema as an eligibility mechanism. Use visible FAQs when they help readers and use structured data accurately, but do not imply that schema guarantees AI visibility.
Conclusion: a preference signal worth testing
Google Preferred Sources creates a new, user-controlled layer in search visibility. Publishers that already have a genuine returning audience should check eligibility, implement the official control on appropriate content, promote it selectively, and measure the results separately from rankings and citations.
For a first snapshot, use the Ranknizer AI Search Visibility tools. When the opportunity requires eligibility review, prompt testing, competitor comparison, and a prioritised implementation plan, consider the $99 manual AI Search Visibility Audit.
Turn a New Visibility Signal Into a Measured Plan
Check eligibility, establish a baseline, and compare preferred-source interactions with AI visibility outcomes.
Explore the Ranknizer tools or request the manual AI visibility audit.
