AI SEO
Google is rolling out natural-language controls that let people tell Discover which topics or links they want to see more or less of. Here is what the change means for users, what publishers should not assume, and the practical content and measurement work that still matters.
By Rahul S., SEO Strategist at SearchCounselCo. Last updated August 25, 2026.
Featured answer: What is Google Discover AI personalization?
Google Discover AI personalization is a new Google app control that lets a person use their own words to request more or less of particular topics or links. Google says the feed will adjust and remember the request. It is a user-level personalization feature not a documented sitewide ranking update or a new publisher markup requirement.
The control
Natural-language requests from a Discover card’s three-dot menu.
The rollout
Announced August 20, 2026 for the Google app, arriving “in the coming days.”
Publisher action
Strengthen topic fit, headlines, images, page experience, and editorial usefulness.
Measurement
Use Search Console Discover impressions, clicks, and CTR without claiming causation.
What Google changed in Discover
On August 20, 2026, Google announced new personalization features across Search, Discover, and Google News. For Discover, Google said users would soon be able to open the three-dot menu on any feed card and describe, in their own words, the topics or links they want to see more or less of. Google says the feed will adjust immediately and remember those requests. The company described the rollout as arriving in the Google app “in the coming days.”
This adds a more expressive control to a feed that already uses a person’s interests and activity. It does not turn Discover into ordinary keyword search. A user is still browsing an interest-based feed; they can now give the system more specific instructions about that feed.
| Question | What changed | What did not change |
|---|---|---|
| User input | People can describe desired or unwanted topics and links in natural language. | Existing follow, unfollow, like, not-interested, and source-blocking controls still exist. |
| Publisher eligibility | No new publisher action was announced for this control. | Indexed, policy-compliant content remains automatically eligible, with no appearance guarantee. |
| Markup | No special Discover personalization schema was documented. | Google still says no special tag or structured data is required for Discover eligibility. |
| Measurement | Publishers have a new event worth annotating in their reporting timeline. | Search Console’s Discover report remains the documented source for page-level impressions, clicks, and CTR. |
Do not confuse this feature with Google’s Preferred Sources button. Current Preferred Sources documentation names Top Stories, AI Overviews, and AI Mode. Discover has its own follow and personalization controls.
How to use the natural-language Discover control
- Open Discover in the Google app. The announced control is tied to the Google app rollout.
- Choose a relevant card. Tap the three-dot menu on a feed item.
- Describe the preference. Ask for more or less of a topic or link in ordinary language—for example, more practical kitchen renovation guides but fewer celebrity-home tours.
- Review the interpretation. Confirm or refine what the interface understood.
- Refresh the feed. Google says Discover will adjust and remember the request.
If the option is missing, the safest explanation is rollout availability. Google announced a phased arrival “in the coming days,” and the help page notes that some Discover controls can vary by language, country, or region. Update the Google app, check again later, and continue using the documented follow, unfollow, not-interested, source-blocking, and like controls. There is no evidence-backed publisher or user hack that forces access.
What AI personalization means for publishers
The most defensible implication is that users can express interest more precisely. If someone asks for “beginner technical SEO fixes for Shopify stores, but not migration advice,” content with a clear subject, audience, and job may be easier for a recommendation system to match than a vague catch-all article. That is an editorial clarity argument, not a claim about a new Google ranking factor.
| User control | Possible publisher implication | Safe action | Do not assume |
|---|---|---|---|
| “More of this topic” | Specific subtopics may receive stronger expressed interest from some users. | Build clearly differentiated topic clusters based on real audience jobs. | A user request guarantees your page will appear. |
| “Less of this topic” | Broad or poorly targeted content may be easier for a user to filter away. | Make titles, images, and opening passages accurate and specific. | A traffic decline proves users rejected your topic. |
| Link-level preference | Particular content examples can help users communicate what they want. | Create useful, distinctive pages worth using as a reference example. | Publishers can see the exact instruction or user identity. |
| Remembered preference | Interest patterns may persist for an individual. | Maintain quality and topic consistency across related articles. | One strong article creates permanent distribution. |
Write for a defined interest, not an imagined prompt database
Google’s current publisher documentation does not describe a report that exposes users’ natural-language Discover instructions. Build content from audience research, first-party search data, customer questions, editorial expertise, and live SERP evidence. Do not manufacture hundreds of thin pages for hypothetical feed prompts. The broader AI SEO fundamentals still start with crawlability, usefulness, clear entities, and verifiable information.
Treat headlines and images as accurate previews
Discover is a visual, interest-based feed. Google recommends headlines that capture the essence of the content and warns against misleading or exaggerated preview details. A compelling image should increase understanding and interest without promising a story the page does not deliver.
Publisher checklist for Discover eligibility and presentation
- Confirm indexability. The canonical page must be indexed and comply with Discover content policies.
- Do not add invented markup. Google says no special tag or structured data is required for Discover eligibility.
- Use a representative large image. Aim for a relevant 16:9 asset at least 1200 pixels wide and more than 300,000 total pixels.
- Enable large previews. Allow
max-image-preview:largeor use an otherwise documented eligible setup. - Declare the preferred image. Use an appropriate
og:imageor relevant schema image property; avoid a generic logo. - Avoid text-heavy artwork. Make the subject legible when the image appears as a small card.
- Write an accurate headline. Capture the page’s real subject without clickbait, withheld context, or sensationalism.
- Add information gain. Timeliness, strong storytelling, or unique insight gives the article a reason to exist.
- Protect page experience. Compress the image, reserve its dimensions, and test the mobile article experience.
- Build a reader journey. Link related guides naturally instead of leaving a Discover landing page isolated.
Use an SEO content calendar to balance fast news responses with durable explainers. Discover can surface both timely and older helpful content, so freshness should not become an excuse for disposable articles.
How to measure the publisher impact without inventing causation
Google’s Discover performance report can show impressions, clicks, and CTR for content that appeared in Discover during the last 16 months, provided the property reaches Google’s minimum impression threshold. A Discover impression is counted when a card is scrolled into view; a click is counted when the user taps the card to visit the result URL. Discover does not use the conventional Search position metric.
If Search Console is unfamiliar, begin with the Google Search Console guide. Keep Discover reporting separate from the dedicated generative AI visibility reporting used for AI Overviews and AI Mode.
A practical 14-day protocol
- Annotate August 20, 2026. Record Google’s announcement separately from the date your audience actually receives the control.
- Freeze a baseline. Export the prior 14 complete days from the Discover Pages report, where available.
- Group pages by theme. Compare topic families rather than celebrating one volatile URL.
- Track impressions, clicks, and CTR. Use the same property, country, device scope, and complete date ranges.
- Log other changes. Publishing volume, major news, internal links, image swaps, indexing changes, and Google updates can all affect the pattern.
- Compare on-site outcomes. Use GA4 landing-page engagement or conversions as a separate post-click layer.
- Write a bounded conclusion. Say “Discover impressions rose after the rollout window” rather than “AI personalization caused the increase.”
Worked example: visibility changed, but the cause is unresolved
The following numbers are fictional. They demonstrate how to interpret the report; they are not SearchCounselCo or client performance data.
| Metric | Previous 14 complete days | Latest 14 complete days | Defensible reading |
|---|---|---|---|
| Discover impressions | 18,400 | 22,600 | Recorded visibility increased by 4,200 impressions. |
| Discover clicks | 740 | 768 | Clicks increased slightly, but much less than impressions. |
| Discover CTR | 4.02% | 3.40% | The added visibility converted to visits at a lower observed rate. |
| Pages receiving impressions | 32 | 41 | Visibility spread to nine additional URLs. |
The safe conclusion is narrow: Discover visibility expanded, clicks rose slightly, CTR fell, and more pages appeared. The data does not show that natural-language personalization caused any of those movements. Check which themes entered the report, whether publishing volume changed, whether a major news cycle affected demand, and whether titles or images changed before choosing the next action.
Google Discover AI personalization FAQ
Is this a Google ranking update?
Google announced a user personalization control, not a sitewide ranking update. The company did not document a new publisher ranking factor, schema type, or opt-in for the feature.
Can publishers see what users type into Discover?
Google’s current public publisher and Search Console documentation does not document access to users’ natural-language instructions. Publishers should use aggregate Discover performance data and their own audience research, not claim prompt-level visibility.
Do I need new schema for Discover AI personalization?
No new schema is documented for the control. Google says indexed content that meets Discover policies is automatically eligible and that no special tags or structured data are required, although eligibility does not guarantee appearance.
Why do I not see the new control?
Google said the feature would arrive in the Google app “in the coming days.” Availability can therefore differ during rollout. Keep the app current and check later; use existing Discover personalization controls in the meantime.
Does Preferred Sources control Google Discover?
Current Preferred Sources documentation names Top Stories, AI Overviews, and AI Mode. Discover has separate follow and personalization mechanisms, so publishers should not present the Preferred Sources button as a documented Discover control.
Can this change guarantee more Discover traffic?
No. Google says Discover traffic is less predictable than keyword-driven search and that eligibility does not guarantee appearance. Measure what changes for your own site without promising a particular result.
The one thing to do next
Export your top Discover landing pages, group them by audience interest, and audit the leading page in each group for headline accuracy, topic specificity, image quality, and a useful next click. Turn the findings into three focused editorial tests—not dozens of speculative prompt pages. If you need the topic map, publishing cadence, and measurement framework designed together, SearchCounselCo’s content strategy service is the relevant route.
Sources used
- Google: Personalize the content you see on Search, Discover, and News — published August 20, 2026.
- Google Search Central: Discover and your website — requirements and measurement guidance checked August 25, 2026.
- Google Search Help: Customize what you find in Discover — existing controls checked August 25, 2026.
- Google Search Console Help: What are impressions, position, and clicks? — Discover counting rules checked August 25, 2026.
The live U.S.-English SERP was observed through a non-personalized, tool-mediated search surface on August 25, 2026. Product controls, rollout, and documentation can change; recheck the linked Google sources before a future material update.