AI Overviews Are Killing Clicks – Now Google Is Backtracking

Wednesday, May 6, 2026, US time. Hema Budaraju, VP of Product Management at Google Search, posts on the official Google Blog: five new updates for AI Mode and AI Overviews. Inline links directly in the text. Hover previews on desktop. A “Subscribed” label for news subscriptions. Article suggestions at the end of the AI answer. Community Perspectives with Reddit and forum quotes.

Sounds like cosmetic adjustments. Google itself frames the updates purely as better web discovery and additional context – the words “click crisis” or “publisher pressure” don’t appear in the official blog post. In my reading, however, the five changes are the clearest reaction yet to the click pressure that independent studies have been documenting for over a year. That’s my interpretation, not a Google statement.

The numbers that forced this step are brutal. Ahrefs measured in February 2026: 58% CTR reduction on the top-ranking page for keywords with AI Overviews. Pew Research documents: 8% click rate with AIO, 15% without – almost half. Chartbeat sees 33% fewer search engine referrals at publishers worldwide. Penske Media (Rolling Stone, Variety, Hollywood Reporter) filed an antitrust lawsuit. The European Publishers Council filed a formal complaint with the EU Commission.

From my work as a Product Developer at iGaming.com and in my SEO Kreativ client projects, I see the pattern on Search Console dashboards: impressions stable or rising, clicks in free fall. That’s “The Great Decoupling” – visibility without traffic.

In this article I break down what Google specifically announced, what the reliable study data actually says (and where outlets are sloppy with citations), why Google is reacting now – and most importantly: what you as a site owner need to do today.

If you want to understand the mechanics behind AI Overviews – Query Fan-Out, Rendered DOM, how Google selects sources – I’ve covered that in my article on how AI Overviews work. This article is about the news, the data behind it, and the actionable takeaways.

The announcement came via the official Google Blog under the title “5 new ways to explore the web with generative AI in Search,” published May 6, 2026. Hema Budaraju, VP Product Management Search, lists five changes – all in rollout simultaneously, but with varying regional availability.

1. Article Suggestions at the End of the AI Answer

Google officially calls this section “Explore new angles.” A new section at the end of AI Overviews and AI Mode answers shows curated links to articles and case studies. Google calls these “relevant article suggestions” – the intent: the AI answer should shift from the endpoint of research to the starting point. Users who have read the answer should be encouraged to click further.

2. Inline Links Directly Next to Relevant Text

Google’s official section title: “See links right where you need them.” Previously, source links primarily sat in a separate source bar or as small icons. Now links appear directly next to the bullet points or sentences they reference. Google’s example: for a search on “California Coast Bike Trip,” a link to a “Pacific coast bike touring guide” sits directly next to the route bullet, while another link to a training blog post sits next to the bullet on daily distance.

3. Hover Previews on Desktop

Google’s official section title: “Get more context on linked websites.” Hovering over an inline link shows a quick preview: site name, page title, occasionally a snippet. Google’s stated goal: reduce the barrier to clicking on unknown domains. This is a micro-optimization at the UX layer.

4. “Subscribed” Label for News Subscriptions

Google’s official section title: “Easily access your news subscriptions.” When the AI answer contains content from a publication the user subscribes to, it’s marked with a “Subscribed” label. Google reports “significantly higher” click rates on labeled links in its own tests. Publishers must actively register for the subscription-linking feature.

5. Community Perspectives

Google’s official section title: “Get advice from people who have been there.” A new section with quotes from public discussions – Reddit, forums, social media – including the creator handle or community name. Google demonstrates this with an example: someone searching “how do I photograph the northern lights” sees quotes from photography forums with exposure tips embedded. Andrew Deck at Nieman Lab notes: this strengthens hobbyist voices over traditional experts – and reduces the click incentive to visit Reddit itself, which Google already licensed for $60M/year in 2024.

From my practice and as a forecast, not as an observed fact: updates 2 (inline links) and 4 (Subscribed) are the two I’d most expect to deliver measurable CTR effects – under the right conditions. Reliable industry data to support that doesn’t exist at the time of the announcement, naturally. Update 5 will radiate to sites with forum/community character. The rest, in my assessment, is window dressing.

Rollout Status in English-Speaking Markets

An important clarification before expectations run too high: Google makes no specific mention of regions, languages, or timelines in the official May 6, 2026 blog post for the five updates. The phrasing is simply “We’re rolling out new updates” – nothing more. This isn’t sloppiness; it’s method: Google rolls out in stages with A/B testing.

What is known and what you can derive from it:

• AI Overviews launched in the US in May 2024 and have since expanded to 100+ countries – the parent function into which updates 1, 2, 3 and 5 feed.
• AI Mode expanded internationally on October 7, 2025, covering 35+ languages and 40+ countries. Both carrier surfaces are live, and the five updates will gradually roll into them.
• The five new updates themselves first appear in English-language US searches from May 6-7, 2026. Sightings in other markets are possible but not a full rollout. Industry observers expect a broader rollout through Q2 2026 – that’s an estimate, not a Google statement.
• Subscription Linking (update 4) requires active publisher registration through Google’s form – regardless of region. Publishers who want the Subscribed label must actively sign up.

What this means practically: in the coming days you’ll see inline links and article suggestions sporadically; a consistent effect will likely set in over weeks. There’s nothing to wait for – the strategic levers in the “What You Should Do” section apply regardless of rollout date.

In media coverage, studies tend to get mixed up. Multiple outlets have written that traffic is down “an average of 58%” according to Ahrefs. That’s imprecise. Ahrefs doesn’t measure “average traffic” – it measures the CTR of the top-ranking page for keywords that trigger an AI Overview. That’s an important distinction and it makes an enormous difference in interpretation.

Pew Research Center – the Methodologically Cleanest Study

The Pew study from July 2025 tracked 900 US adults who voluntarily shared their browsing data – across 68,879 real Google searches in March 2025. Result: when an AI Overview appears, only 8% of users click on a traditional search result. Without an AI Overview, it’s 15%. That’s a relative decline in click rate of roughly 47% – not a statement about total traffic volume, but about per-session click probability. Clicks on links within the AI answer itself: 1%. And: 26% of sessions end directly after an AIO page, compared to 16% without.

The study is particularly valuable methodologically because it measures actual behavior – not just SERP tracking. Google itself labeled the methodology as “flawed.” Pew stands by the numbers. So do I – browsing data methodology is exactly what pure SERP analysis misses.

Ahrefs – the Largest Keyword Sample

Ryan Law (Director of Content Marketing) and Xibeijia Guan (Data Scientist) in the updated Ahrefs study from February 4, 2026 analyzed 300,000 keywords – 150,000 with AIO, 150,000 without. Comparison period: December 2023 (before AIO rollout) vs. December 2025. Method: aggregated Google Search Console data on average desktop CTR per keyword.

The result after correcting for general SERP trends: 58% lower CTR on the top-ranking page when an AI Overview is present. In April 2025, the same methodology showed 34.5% – the effect intensifies as the AIO rollout matures. Important: this figure applies to the top-ranking page for AIO keywords. It is not “the average traffic loss across all sites.”

Seer Interactive – the Longest Time Series

Tracy McDonald at Seer Interactive has been tracking since June 2024 – 3,119 informational keywords across 42 organizations with a total of 25.1 million organic and 1.1 million paid impressions. The data is clear: in the June 2024 baseline, organic CTR for AIO keywords was 1.76%. By September 2025, it fell to 0.61%. A 65% decline for AIO keywords. For non-AIO queries over the same period, CTR fell “only” from 2.74% to 1.62%.

In the April 2026 update, Seer documents a tentative recovery: organic CTR for AIO queries climbed from 1.3% in December 2025 to 2.4% in February 2026. Seer itself calls this “leveling off” – not a real recovery, but stabilization. The floor appears to have been reached; Q4 2025 was the outlier to the downside.

What the Numbers Show Together

Three different methods – browsing data (Pew), keyword CTR (Ahrefs), brand-level GSC tracking (Seer) – are not directly interchangeable, but they show the same trend in different measurement contexts. Reading all three side by side: the click losses are real, they are large, and they are methodologically confirmed by multiple independent actors. The magnitude ranges from 30% to 90% depending on content type, position, and measurement method. The direction is unambiguous – the exact number depends on what precisely is being measured.

Comparison Table of Key Studies

From my work in technical SEO audits, I see this pattern in almost every GSC account serving informational queries: impressions hold, clicks erode. Anyone who hasn’t noticed anything in their Search Console data yet either primarily handles transactional or local queries – or is looking at the numbers too coarsely.

The five updates don’t come out of nowhere. They are the result of concentrated pressure that has built massively over the past 12 months.

Pressure Point 1: Antitrust Lawsuits

In September 2025, Penske Media Corporation – publisher of Rolling Stone, Variety, Deadline, The Hollywood Reporter, and Billboard – filed an antitrust lawsuit against Google. The allegation: abuse of search monopoly position to compel publishers to accept AI training and AIO utilization. In February 2026, Penske filed a memorandum against Google’s motion to dismiss – with detailed documentation of traffic losses. The European Publishers Council filed a formal complaint with the EU Commission.

Pressure Point 2: Looming Publisher Exodus

A Search Engine Land survey of 350+ SEOs found: 33.2% want to block Google once opt-out mechanisms are available, 24.9% are undecided. The UK’s CMA (Competition and Markets Authority) ran a consultation phase that ended in February 2026. Industry observers expect opt-out tools by mid/late 2026 – and Google risks CMA enforcement if it doesn’t deliver.

Pressure Point 3: Self-Interest

This is where it gets interesting. In the official Alphabet Q1 2026 quarterly report from April 29, 2026, Google Search & Other reports $60.4 billion in revenue – a 19% increase year-over-year. Google’s total ad business: $77.25 billion in a single quarter. Sundar Pichai on the earnings call: “People love our AI experiences like AI Mode and AI Overviews, and they’re coming back to Search more.” These revenues depend on the web continuing to produce content and advertisers continuing to want to appear on search result pages.

Noteworthy: a detail discrepancy in the same quarterly report. Google Network (external AdSense partner sites) fell 4% to $6.97 billion, while Search on Google’s own SERPs grew 19%. More queries processed within Google’s own interface generate fewer external clicks – the mechanism depresses Network revenues while overall search revenues rise. The publishers’ click crisis is measurable on Google’s own balance sheet.

If publishers die off from lack of traffic monetization or opt out, the raw material for AI answers disappears long-term – and with it, the advertising inventory. The five updates are the attempt to manage the trade-off: keep AIO as the primary interface while letting just enough clicks through that the ecosystem doesn’t collapse.

The updates are arriving with significant PR echo. The reality is more sober. Three aspects you shouldn’t lose sight of:

What Can Realistically Help

Inline links next to bullet points have structurally more click potential than links in a side source bar. Hover previews lower the barrier because users see where the click goes. The “Subscribed” label with active subscriptions is actually a real UX boost – Google itself reports “significantly higher” clicks in tests. For news publishers with a working subscription setup, that’s a direct plus.

What Is Window Dressing

Article suggestions at the end are rarely the click trigger in practice. Anyone who has read to the end of a comprehensive AI answer has usually had their question answered. Community Perspectives (Reddit quotes, etc.) reduce the incentive to visit Reddit itself – the click moves into Google’s ecosystem. The structural problem remains: users have the answer by the time they reach the source.

What Doesn’t Heal the Structural Wound

The fundamental mechanic remains: a complete AI answer makes the click optional. Pew shows that only 1% click on links within the AIO answer. When the answer is complete, the source attribution is a consolation prize, not a traffic driver. The five updates improve the consolation prize. They don’t make it the primary traffic artery.

In my current client projects I observe: sites cited in AIOs benefit substantially. Seer documents in the April 2026 update “+120% more organic clicks per impression” for cited brands compared to non-cited ones – and a consistent advantage of roughly 4 percentage points in paid search CTR throughout all of 2025. Citation optimization instead of classic position-1 hunting is the only lever that holds up reliably.

1. Use Search Console as an Analysis Lever

Google added dedicated AI Search tracking options to GSC in mid-2025. Under Performance on Search Results, filter by “Search Appearance” – you now get “AI Mode” and “AI Overviews” as separate segments. Use them to segment impressions, clicks, and CTR. Compare values per segment. Identify the pages that have AIO impressions but almost no clicks. That’s your citation pipeline – and simultaneously your traffic-loss hotspot. The filter isn’t a cure-all, but it’s currently the most precise diagnostic layer for AIO effects directly in your own data.

2. Build Citation Readiness (One Lever, Not the Lever)

The Seer data is clear: brands cited in AIOs get +120% more organic clicks per impression than non-cited sources according to the April 2026 update – with a consistent paid search advantage of roughly 4 percentage points throughout all of 2025. That doesn’t mean “citation replaces everything.” It means: citation readiness is one lever that shows particularly measurable impact. Concretely: clear answer blocks (answer-first principle, 50-70 word direct answer at the start), clean structure with lists and tables, strong entities and consistent author profiles, reliable sources, and clean schema markup. Only the combination increases the probability of appearing as a citable source in AI answers.

3. Structured Data, Consistently

Schema.org for the central entities: Article, FAQPage, HowTo, Product, Organization, Person. The cleaner the semantic markup, the higher the probability of AI pickup. Validate via the Rich Results Test and Schema Markup Validator – broken markup actively hurts.

4. Make E-E-A-T Signals Visible

Author profiles with demonstrable expertise. Original data and proprietary studies. Qualified backlinks. AI systems evaluate authority not just on your site, but across your entire digital footprint – social, forums, reviews. After the March 2026 Core Update, E-E-A-T is even more important than before.

5. Actively Track Brand Search

Isolate searches containing your brand name in GSC (filter: query contains brand name). Brand search is the resilience factor par excellence against every future AIO update. Those with a brand survive. Those serving only generic informational queries are commodities – and therefore replaceable by AI synthesis.

6. Search Everywhere Optimization

Stop relying on Google alone. Reddit (37%), YouTube (19.8%), Quora (8%), LinkedIn (5%) are now the most visible platforms in Google SERPs themselves according to SE Ranking. Building presence there captures traffic on two levels: directly through the platform and indirectly through Google visibility.

I read the May 6, 2026 announcement as the clearest indirect reaction Google has given yet to the documented click pressure – not explicitly acknowledged in the official blog post, but the logic is unmistakable. But it doesn’t solve the fundamental problem. It can’t solve it methodologically – as long as the AI answer is complete, the click remains optional. The updates improve the consolation prize. They don’t redefine the SERP.

Those who want to remain resilient over the next 12 months shouldn’t optimize for these five features. Instead, optimize for the logic behind them: citation matters, brand matters, structure matters. Position 1 is no longer the goal – being the source of an AI answer is. And that is a fundamentally different discipline than classical SEO.

From my work as a Product Developer and SEO consultant: the sites that are climbing slightly again in the April Seer update are exactly those that started this pivot 6 to 12 months ago. The others are still waiting. Waiting is the more expensive strategy.