Google Adds AI Search Links, Still Hides Click Data

Google AI search links got a visible upgrade in May 2026, with AI short for artificial intelligence, the systems that generate and source answers. The company added inline citations, subscription labels, forum perspectives, desktop previews and a new end-of-answer article module across AI Overviews and AI Mode, while Search Console still folds AI clicks into ordinary Web search reporting.

That makes the update less of a peace offering than a measurement test. Links can help only where the answer leaves enough curiosity to click, and publishers still cannot separate a better link from a lost visit in Google’s own reporting.

Five Link Surfaces, One Reporting Gap

Hema Budaraju, Google vice president of product management for Search, presented the changes in Google’s May product note on AI search links. The package has five new places to click, but none changes the reporting layer that publishers use to prove whether those clicks arrived.

The distinction matters because each surface solves a different user problem. Some reduce friction. Some add context. One favors paying subscribers. The publisher question is narrower: can the site owner see the incremental traffic by feature, query and page?

Search Console Keeps AI Clicks in the Web Bucket

The hard limit sits in Google’s own documentation. In Google’s AI Search Console guidance for site owners, the company says appearances in AI features are included in overall Search Console traffic and reported inside the Web search type. That gives publishers totals, not a clean split.

Search Console remains the bottleneck because it is the shared ledger for search engine optimization, editorial planning and ad sales. A news site can see that a page gained impressions and lost clicks. It cannot tell from the native report whether an inline citation helped, a generated answer cannibalized the query, or a subscription label lifted paid-reader traffic.

Google also keeps arguing that clicks from AI-enhanced search are higher quality. The company defines that in its documentation as users being more likely to spend time on a site. That is useful if a publisher can connect the click to a session in analytics. It still leaves the first question unanswered: how many AI-surfaced clicks arrived in the first place?

Independent Data Keeps Pressing the Same Point

Outside measurements have filled the vacuum, and they keep pointing in the same direction. Pew Research Center’s Google AI summary study used browsing data from 900 U.S. adults and 68,879 unique Google searches to compare behavior with and without AI summaries.

The newer academic work adds a second problem beyond traffic. The arXiv study on AI Overview activation and publisher impact found that 11.0% of the atomic claims it checked were unsupported by the cited pages. A link beside a claim does not guarantee that the cited page backs the claim cleanly.

That matters for publishers because the link can become reputational risk without guaranteed traffic. A site may get cited, may not get clicked, and may still be associated with a generated claim that does not match its article.

The best version of Google’s May change would make source proximity more honest and more useful. The unresolved version makes citations more visible while the click counter stays blended.

Subscription Labels Move the Bargain Closer to Google

Subscription labels are the sharpest part of the package for paid publishers. Google says early testing made users significantly more likely to click labeled links, but it did not publish the size of that lift. The company also asks publishers to submit information so it can connect a subscriber’s paid sources to the search experience.

• Paid news brands get a new way to remind subscribers that they already have access.
• The label appears inside a Google-controlled AI surface, not inside the publisher’s own product.
• The publisher must judge the lift without a native feature-level report in Search Console.

That is why the feature feels useful and uncomfortable at the same time. A new Google-controlled label may help a subscriber choose a paid source over a generic answer. It also makes Google part of the paid-reader habit loop that publishers have spent years trying to move back to their own apps, newsletters and homepages.

Regulators Are Asking for Controls

The United Kingdom’s Competition and Markets Authority, the country’s competition regulator, has already moved past link placement. Its January 2026 proposals for Google Search call for publisher controls, fair ranking, choice screens and data portability. The CMA package for Google search services says publishers should get more choice and transparency over how their content is used in AI answers.

That regulatory framing is important. It treats attribution as one piece of a larger bargain, not the full repair. The issue is not only whether a link appears. It is whether a publisher can decline certain AI uses without losing ordinary search visibility, see how ranking works, and measure the effect on traffic.

The European Commission, the executive arm of the European Union, is pushing on a related front through the Digital Markets Act (DMA, the EU rulebook for large online gatekeepers). In the U.S., Penske Media Corporation, publisher of Rolling Stone and The Hollywood Reporter, has challenged Google’s AI search practices in federal court. The pressure is converging on the same gap: choice and transparency need operational controls, not just nicer citations.

Revenue Has Moved Faster Than Measurement

Alphabet’s search business has not shown the pain that publishers describe. In its first quarter Alphabet earnings release, the company reported Google Search and other revenue of $60.399 billion for the quarter ended March 31, 2026, up from $50.702 billion a year earlier. Sundar Pichai, Alphabet and Google chief executive, also said queries reached an all-time high.

Google Network, the ad business that includes ads placed on partner properties, moved the other way in the same table, falling to $6.971 billion from $7.256 billion. That line does not prove AI search caused the decline. It does show why publishers watch the mix closely: search ads can grow while partner-site economics weaken.

Chartbeat, the publisher analytics company, has reported that chatbot referrals remain less than 1% of pageviews across its network, even as ChatGPT and Perplexity traffic grows. That means replacement traffic from AI assistants is still too small for most publishers to offset losses from ordinary search referrals.

If Google creates separate filters for AI Search surfaces, the May link changes become measurable. If it keeps reporting them as Web search, publishers will be left comparing screenshots, rankings and server logs while the most important click counter stays blended.