Google 测试 Discover「Dive deeper」AI 概览，引发出版商流量担忧

Google is testing a new Dive deeper experience in Google Discover that inserts an AI-generated topic overview between a Discover card and the publisher content a user might otherwise open. The experiment begins with video cards, but its significance extends beyond video: Discover has been an important referral channel for publishers, and a Google-hosted summary could change where user attention goes before a site receives a click. Google Search VP Robby Stein announced the Discover test publicly on September 18, 2026. He said that tapping “Dive deeper” opens a short topic overview with prominent links to related stories, community reactions, and original reporting. Google plans to test multiple designs in the coming weeks. This is a live experiment, not a confirmed wider rollout, and Google has not disclosed countries, languages, traffic results, rollout dates, or an opt-out mechanism. The immediate concern for publishers is not that links disappear. The illustrated experience still includes publisher coverage. Instead, the change introduces a Google-generated first destination that may satisfy some users before they visit an external article. That makes the design relevant to any publisher that depends on Discover referrals for audience growth, advertising impressions, subscriptions, affiliate revenue, or lead generation. What Google’s Discover test changes The test changes the sequence of actions available to a Discover user. Rather than moving directly from a card to a publisher page, a user can first open an overview designed to provide context around a topic. Reporting on the test describes an initial AI-generated summary, followed by a screen containing related-source cards and an “Explore more stories” area. Aspect Direct Discover journey Dive deeper test First destination after engagement A publisher’s selected card or destination A Google-generated topic overview Content initially covered Discover cards broadly Video cards at the start of the test Publisher visibility Publisher content is the primary card destination Related stories and original reporting links remain in the experience Scope Established Discover behavior Experiment, with multiple designs planned This distinction matters because a link can remain visible while receiving a different share of clicks. A concise overview may encourage a user to continue exploring Google’s presentation of the topic, or it may make the subsequent publisher links more useful by giving them context. Google has released no referral or click-through data, so neither outcome can yet be quantified. The test also fits Google’s broader use of AI-generated answers and summaries in Search products, including AI Overviews and AI Mode. Discover is different from a conventional search results page because it is a recommendation feed. Publishers cannot assume that Search behavior, referral patterns, or optimization tactics will transfer directly to this surface. Why the initial video focus still matters A video-first test narrows the immediate scope, but it does not remove the need to watch it. Video cards can be a meaningful distribution surface for publishers and creators, while the test gives Google a way to measure how people use topic exploration inside Discover. PPC Land reported that independent researchers observed code flags that may point to broader formats, including article cards. Those flags are not a confirmed expansion. Google’s public statement does not say that Dive deeper will move beyond video cards, and publishers should avoid treating a potential article-card rollout as settled fact. What publishers should monitor and do now For now, the practical response is measurement and distribution resilience, not a wholesale content overhaul. Publishers should establish a baseline for Discover performance before any change reaches their audience. That makes it easier to identify a real shift in referrals instead of mistaking normal Discover volatility for the effect of a new interface. Useful signals to track include: Discover clicks and impressions in Search Console, segmented by content type where internal analytics allows. Landing-page engagement, including returning users, pages per visit, subscription starts, leads, and ad-view depth from Discover referrals. Video performance, especially for sites that receive Discover traffic through video cards. Topic-level patterns, such as whether explanatory or breaking-news content gains or loses visibility over time. Referral concentration, so teams know how exposed they are if Discover traffic declines or becomes less predictable. Content quality remains important because the test explicitly points users toward related stories and original reporting. A generic summary may answer basic questions, but it is less likely to replace reporting that adds firsthand information, distinctive expertise, useful evidence, clear visuals, or a genuinely local perspective. Publishers can make this value easier to recognize by using precise headlines, strong lead paragraphs, clear author and publication information, and pages that deliver the specific promise made by the Discover card. Distribution is the other contingency. Businesses that publish to generate revenue should avoid making Discover their only acquisition path. Strengthening newsletters, direct audience relationships, social distribution, syndication where appropriate, and repeat visits can reduce exposure to changes in a single Google surface. This is not evidence that Discover traffic will fall. It is a practical response to the fact that the experiment creates another point where a user may remain within Google before reaching a publisher. For publishers, the key question is not simply whether AI summaries appear. It is whether the placement, prominence, and click paths of publisher links preserve meaningful referral value. Google’s planned design iteration means those details may change before any broader deployment. If Discover is a material source of leads, subscriptions, or advertising revenue, Scalevise can help turn scattered analytics into a practical content-distribution plan. Our AI Visibility and GEO Checker can help identify where your brand and content appear in AI-driven discovery experiences, while clarifying the pages and topics that deserve closer measurement. Build a clearer baseline now, reduce dependence on one referral channel, and start an AI Visibility scan. Frequently Asked Questions What is Google Discover’s Dive deeper test? Dive deeper is a live Google Discover experiment. When a user taps the option, Google opens a short AI-generated topic overview with links to related stories, community reactions, and original reporting. Does Dive deeper replace publisher links in Google Discover? No. Google’s example includes links to publisher coverage. However, the AI topic overview becomes an additional step before a user may select an external publisher link. Which Discover content is included in the test? Google said the test starts on video cards. Google has not confirmed expansion to article cards or other Discover formats. Has Google published traffic-impact data for Dive deeper? No. Google has not published referral, click-through, or publisher-traffic data for the experiment. What should publishers watch first? Publishers should monitor Discover clicks and impressions, video performance, landing-page engagement, and the share of their traffic that depends on Discover referrals. Conclusion Google’s Dive deeper test places an AI-generated topic layer inside Discover while retaining links to related publisher coverage. Its eventual effect on publisher traffic remains unknown, but the experiment makes referral measurement, distinctive reporting, and diversified distribution more important. The details to watch are whether the test expands beyond video cards and how prominently Google continues to surface external publisher links.