To win citation prominence in Google's AI results, run two separate playbooks. To become a Preferred Source, earn the brand affinity that makes real people add you in their settings: direct traffic, a newsletter, and a name worth typing. To earn a "Highly Cited" label, become the source other publications cite by publishing original research, primary reporting, and proprietary data behind clean, expert-authored, well-structured pages. They are different levers, and only one of them is something you can engineer on-page.
- EffectivePreferred Sources expanded to AI Overviews and AI Mode on May 27, 2026
- AdoptionMore than 345,000 unique sources selected at launch
- ClickthroughPeople are twice as likely to click a Preferred Source
- The modelAI Mode now runs on Gemini 3.5 Flash as the default
- The mechanicQuery fan-out cites passages, not whole pages
- The splitPreferred Source is brand work, Highly Cited is content and PR
Two prizes opened up in Google's AI results, and they reward opposite kinds of work. A Preferred Source is a site a user personally chose to see more of, so it is earned through brand affinity you cannot directly engineer. A "Highly Cited" label is a badge attached to articles many other stories reference, so it is earned through original reporting and data that you absolutely can engineer. Treating them as one goal is the mistake. This guide separates them, gives you the distinct playbook for each, and ends with a working checklist and a measurement plan.
CH.01Citation is not ranking - understand the surface first
On May 27, 2026, shortly after Google I/O, Google expanded its Preferred Sources feature into both AI Overviews and AI Mode, letting users designate sites they want surfaced more prominently in AI results. The same rollout added prominent article carousels for developing-topic queries and "Highly Cited" labels on links that are widely cited by other publications. AI Mode is now the default and runs on Gemini 3.5 Flash, so being cited is a distinct objective from classic ranking.
Before either playbook makes sense, internalize how the AI surface selects what it shows. Classic ranking returns ten blue links ordered by relevance to a single query. Google's AI experiences work differently, and the mechanics change what wins. Google uses a technique it calls query fan-out, issuing multiple related searches across subtopics and data sources to assemble an answer. The model then stitches together passages from many pages rather than promoting one whole document.
Google states there are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible for a snippet. Eligibility gets you considered; the two playbooks below are what get you chosen and labeled.
Three consequences of a passage-based, fan-out surface
- You compete passage by passage, not page by page. A single self-contained section that cleanly answers one subtopic can be cited even if the full page does not rank first for the head term.
- Coverage breadth matters more than a single hero page. Because fan-out explodes one query into many, sites that answer the full constellation of related sub-questions get pulled in more often.
- Being eligible is the floor, not the ceiling. Indexing and snippet eligibility get you considered, and the two playbooks decide whether you are chosen and labeled.
This is why AI visibility is its own workstream. For the broader picture of operating in an AI-Mode-default world, see our breakdown of what AI Mode as the default means for search visibility, and for the traffic-economics reality, our analysis of zero-click search and publisher traffic in 2026.
CH.02Playbook 1 - Becoming a Preferred Source
A Preferred Source is a site a user has personally added in their Google settings so it gets surfaced more prominently in their AI results. The selection is user-controlled, which is the single most important fact about this feature: you cannot directly game it. You earn it by becoming a brand people want to hear from.
How the feature actually works
Users add Preferred Sources at google.com/preferences/source, choosing which sites should be highlighted in their results. When a chosen site appears in an AI Overview or AI Mode response, its link is clearly labeled to stand out. According to 9to5Google, Google also plans to use these selections as a ranking signal so preferred sites appear more frequently. Treat that ranking-signal detail as press-reported emphasis rather than fully specified mechanics.
Per Google, the rollout began May 27, 2026, more than 345,000 unique sources had already been selected at launch, and people are twice as likely to click through to a Preferred Source.
The strategic implication is blunt. The lever is not a meta tag or a schema property. It is whether a real person, sitting in their settings, thinks of you by name and decides they want more of you. That is a brand-and-audience problem, so solve it like one.
How to earn the action that gets you added
- Build a brand worth typingPreferred Sources rewards name recognition. Invest in a distinctive editorial point of view, a recognizable masthead or author roster, and consistent quality so readers associate your domain with a category. Generic, interchangeable content never gets manually selected.
- Convert readers to direct relationshipsA newsletter is the highest-leverage asset here. Email subscribers already chose to hear from you, so they are your most likely Preferred-Source adopters. Pair it with RSS, app or push subscriptions, and an obvious follow path.
- Drive direct and branded trafficPeople add sources they navigate to on purpose. Branded search demand, direct visits, and a memorable domain all signal the loyal readership that translates into a manual "make this a Preferred Source" action. Track branded query volume and direct sessions as leading indicators.
- Prompt the action in your own channelsOnce the feature is live for your audience, tell readers how to add you. A short evergreen explainer, a footer note, and social reminders are fair game. You are not gaming the algorithm; you are asking loyal readers to use a setting Google built for exactly this.
- Be consistently useful, not occasionally viralReliability beats spikes. Someone adds a source because they expect future value, so a steady cadence of genuinely helpful, people-first content sustains the relationship the setting formalizes.
CH.03Playbook 2 - Earning "Highly Cited" labels and carousel placement
The second prize is more engineerable. A "Highly Cited" label is a badge Google attaches to article links that many other stories have cited, helping users find the primary reporting that other articles reference. Google also surfaces a prominent article carousel for developing-topic queries, and a separate perspectives carousel that pulls insights from online discussions, forums, and social media. To win these, you have to be the source other people cite. That is a content and digital-PR problem with a clear formula.
Become the thing others cite: originality and primary data
The label is awarded for being referenced by other publications, so produce material that other publications have a reason to reference.
- Publish original research and proprietary data. Survey your audience, analyze your own platform data, or run an experiment, then publish the numbers. A citable statistic is the single most linkable asset on the web because writers need a source for claims.
- Do primary reporting. Break news, interview named experts, attend the event, file the firsthand account. Synthesis of other people's reporting does not earn citations; original reporting does.
- Package data to be quoted. Give each key finding a clean, self-contained sentence with the number stated explicitly, a clear methodology note, and a chart with a citable caption. Make it trivial for a journalist to lift one stat and link back.
- Coin and define. Naming a framework, index, or term that others adopt creates a durable citation magnet, because every future mention points to your definition.
A citable statistic is the single most linkable asset on the web. If a finding cannot be lifted as one self-contained, attributable sentence with the number stated explicitly, it will not become the thing other publications cite. Capconvert GEO practice
Amplify it: digital PR and developing-topic speed
- Run digital PR around your dataPitch your original research to journalists and trade publications. Citations from many independent publications are precisely the signal the "Highly Cited" label measures, so concentrate effort on assets built to be referenced.
- Cover developing topics fast and wellThe developing-topic carousel rewards timely, substantive coverage. When a story breaks in your category, publish a clear, accurate, well-sourced piece quickly, then keep it updated. Speed plus quality lands carousel placement and accumulates early citations.
- Earn community perspective placementThe separate perspectives carousel surfaces forums and social discussion, building on the May 6, 2026 link updates that added firsthand perspectives to AI responses. Genuine, helpful participation under named experts feeds this surface.
Make it citable: E-E-A-T, authorship, and structure
Originality earns the citation; trust and structure make Google confident enough to label and surface it.
- Put expert authorship front and center: real, credentialed bylines with author bio pages and links to the author's body of work.
- Demonstrate first-hand experience with original photography, raw data, "we tested this" detail, and specifics only a practitioner would know.
- Structure for passage-level extraction: a direct answer in the first sentence under each heading, short paragraphs, descriptive subheads, and lists for parallel facts.
- Use structured data where it fits: Article with author, datePublished, and dateModified; FAQPage where genuinely applicable; Organization markup.
- Keep the page clean and crawlable: available in textual form, crawlable, and internally linked so passages can be extracted.
Because AI assembles answers from passages, a self-contained section that stands alone when quoted is a section that gets cited. Anonymous content rarely earns a primary-reporting label, and important facts buried in images or behind scripts cannot be extracted at all.
For how this connects to the broader shift toward AI agents doing the browsing on a user's behalf, see our guide to staying visible when Google's AI agents do the browsing. The whole I/O 2026 picture is in our complete breakdown for search marketers.
CH.04The concrete checklist
Use this as a working punch list across both playbooks. The first list is audience and brand work for Preferred Sources; the second is content and PR work for Highly Cited labels and carousels.
Preferred Sources (audience and brand)
- Launch or grow a newsletter and make subscription prominent on every page.
- Track branded search volume and direct traffic as your Preferred-Source leading indicators.
- Develop a distinctive editorial voice and a recognizable author roster.
- Publish a short evergreen explainer telling loyal readers how to add you at google.com/preferences/source, plus a footer reminder.
- Maintain a reliable publishing cadence so readers expect future value.
Highly Cited and carousels (content and PR)
- Ship at least one original-research or proprietary-data study per quarter, with a clear methodology.
- Package every key finding as a quotable, self-contained statistic with a citable chart.
- Pitch that research to journalists and trade publications through active digital PR.
- Stand up a rapid-response workflow to cover developing topics in your category within hours, accurately.
- Use real expert bylines with full author bio pages on every substantive piece.
- Write passage-first: a direct answer in sentence one under each heading, short paragraphs, and lists.
- Implement Article, Organization, and (where genuine) FAQPage structured data, with accurate datePublished and dateModified.
- Confirm content is textual, crawlable, and internally linked so passages can be extracted.
CH.05How to measure progress
You cannot see who added you as a Preferred Source, so measure proxies. For Playbook 1, watch newsletter growth, direct and branded-search traffic, and returning-visitor rate. For Playbook 2, track how often other publications cite or link to your original-data pages, monitor which of your URLs appear in AI Overviews and AI Mode citations, and watch for the "Highly Cited" badge and carousel placement on your developing-topic coverage.
Set up alerts for unlinked brand mentions and new referring domains around each data study. A rising citation count is the leading indicator that the "Highly Cited" label is within reach.
From here the loop repeats. Build the brand and the direct relationships that earn Preferred-Source selection, ship the original data and primary reporting that earns citations, structure every page for passage-level extraction, then measure citation count and AI presence rather than a single keyword rank.
FAQCommon questions
Can I make my site a Preferred Source through SEO?
No. Preferred Sources is a user-controlled setting: a person manually adds your site at google.com/preferences/source so it is surfaced more prominently and clearly labeled in their AI results. There is no meta tag, schema property, or technical optimization that designates you. You earn it indirectly by building brand affinity, a newsletter, branded demand, and direct traffic, then prompting loyal readers to add you. Treat it as audience development, not on-page SEO.
What does the "Highly Cited" label actually mean?
A "Highly Cited" label is a badge Google adds to article links that many other stories have cited, helping users find the primary reporting other articles reference. Per Google's May 27, 2026 announcement, it also flags when an article explicitly references a Highly Cited source. To earn it, publish original reporting, proprietary research, and citable data that other publications have a reason to link to, then amplify those assets with digital PR.
How is being cited in AI results different from ranking?
Classic ranking orders whole pages for one query. Google's AI results use query fan-out, issuing many related searches and assembling an answer from passages across multiple pages. So you compete passage by passage, and breadth of coverage across related sub-questions matters more than a single top-ranked page. A self-contained section answering one subtopic can be cited even when the full page is not ranking first for the head term.
Does AI Mode running on Gemini 3.5 Flash change what I should do?
Not the fundamentals. Google upgraded AI Mode to Gemini 3.5 Flash as the new default model globally, which improves answer quality and agentic capability, but the inputs are the same: indexed, crawlable, helpful, people-first content available in textual form. Model upgrades change how answers are generated, not what makes a source trustworthy and extractable. Keep optimizing for clarity, originality, and E-E-A-T.
What is the developing-topics article carousel and how do I get into it?
It is a prominent carousel Google shows for searches about evolving stories, surfacing timely articles so readers can get initial context and choose where to explore. A companion carousel surfaces perspectives from forums and social discussions. To earn placement, cover breaking topics in your category quickly and accurately with expert bylines and clear sourcing, then keep the piece updated. Speed combined with substantive, citable quality is what wins carousel slots and early citations.
Should I add structured data to get cited in AI Overviews?
Google states no special schema is required to appear in AI Overviews or AI Mode; standard indexing and snippet eligibility suffice. That said, accurate Article, Organization, and genuinely applicable FAQPage markup clarifies authorship, publisher, and dates for machines, which supports the trust and freshness signals behind labels like "Highly Cited." Use schema for accuracy and clarity, never as a substitute for original, well-authored content.
References
- Google. "Helping you find original, high-quality content in Search." blog.google/products-and-platforms/products/search/original-high-quality-content-search
- Google. "What's new with Search at Google I/O 2026." blog.google/products-and-platforms/products/search/search-io-2026
- Google Search Central. "AI features and your website." developers.google.com/search/docs/appearance/ai-features
- Google Search Central. "Creating helpful, reliable, people-first content." developers.google.com/search/docs/fundamentals/creating-helpful-content
- 9to5Google. "Google AI Mode gets Preferred Sources." 9to5google.com/2026/05/27/google-ai-mode-preferred-sources
- 9to5Google. "Google AI Mode and Overviews add direct links." 9to5google.com/2026/05/06/google-ai-mode-overviews-direct-links
- Schema.org. "Article." schema.org/Article