Google hasn't simply added an answer box to Search. It has moved more of the web's work onto Google's own page, and publishers are now fighting over the clicks that used to pay for that work.
The current Google Search story isn't a clean July 10 switch that made every query an AI answer. That claim doesn't hold up. The real story is narrower, and more useful: AI Overviews and AI Mode have become big enough that anyone who depends on search traffic now has to treat Google as both distributor and competitor.
That is the break.
Google introduced Gemini 3.5 Flash at I/O on May 19, 2026, and Google's own developer notes say the model was released that day as the generally available version behind gemini-flash-latest. Google also said in May that AI Mode had passed 1 billion monthly active users globally, while AI Mode queries had more than doubled every quarter since launch. You don't need to pretend every search is now AI-generated to see the problem. The feature is already operating at a scale where a small change in user behavior can move real money away from publishers, affiliates, and startups that built their acquisition plans around organic search.
The numbers are blunt. Ahrefs said in February 2026 that it re-ran its study across 300,000 keywords using December 2025 Search Console data and found AI Overviews correlated with a 58% lower click-through rate for the top-ranking page. Pew Research Center found in July 2025 that Google users clicked a traditional search result in 8% of visits when an AI summary appeared, compared with 15% when it did not. Pew also found users clicked links inside the AI summary in just 1% of visits. That's it.
If you run a media site or an SEO-dependent SaaS funnel, that assumption just broke. Organic search used to mean ranking well enough that a stranger had to visit your page to get the answer. Now the answer may appear before your link gets a serious look. Google still points to sources, but a citation that nobody clicks is not distribution. It is decoration.
The court fight is no longer theoretical
Regulators and courts are starting to treat AI search differently from old search. A Munich regional court issued a temporary injunction on May 28, 2026, in case 26 O 869/26, after Google's AI Overview falsely associated two Munich-based publishers with scams and subscription traps. The Library of Congress's legal monitor summarized the ruling this month, and several German legal analyses reached the same point: the court treated the AI Overview as Google's own generated statement, not merely a neutral list of third-party links.
That distinction matters. A normal search engine can say it is surfacing other people's pages. An AI Overview rewrites, selects, and combines material into a new answer. Once Google does that, it becomes harder to argue that the company is only showing where information lives. It is speaking in its own format and on its own page - above the material it summarizes.
Germany's media regulator moved in the same direction on July 14, 2026. Reuters reported that the Commission for Licensing and Supervision, known as ZAK, said Google's AI Overviews and Perplexity's AI search fall under German media law because AI-generated summaries and chatbot responses are content created by the providers themselves. That is not a small bureaucratic footnote. It is a direct challenge to the idea that AI search can enjoy the reach of publishing without the duties that come with publishing.
Frankly, that is where Google is vulnerable. If an AI answer keeps users on Google, sells against their attention, and shapes what they believe, it will not be treated forever as a passive index. Courts may move slowly, but they have found the weak spot.
Founders need another route to the reader
None of this means you should abandon search. It means you should stop treating it as owned distribution. There is a difference. Search can still bring customers, investors, readers, and buyers to your business, but the old playbook of publishing a helpful page and waiting for Google to rank it, then hoping the visitor converts on your own site, is less reliable than it was.
The content that survives this shift will be more specific, not broader. Models cite clear facts, named products, original data, direct comparisons, prices, dates, and documents they can parse. Thin explainers are easy for AI search to absorb and replace. A pricing table, a benchmark, a founder interview, a primary dataset, or a technical teardown is harder to flatten into a generic answer without leaving the user wanting the source.
You also need channels Google can't interrupt. Email still looks dull until you need it. So do paid communities, direct sales lists, podcasts with loyal audiences, customer Slack groups, and product-led loops that bring users back without a search box in the middle. Some publishers are trying licensing deals with AI companies. Others are leaning into apps and newsletters. Neither path fixes the entire traffic problem, but waiting for Google to send the old clicks back is not a plan.
The legal fights in Munich and Brussels will not resolve fast enough for the businesses being squeezed right now. Google has the audience, the ad stack, and the interface. Publishers and founders have the underlying work. That bargain can still function, but only if the people making the work get something more durable than a citation buried inside an answer. Until then, build as if the traffic is not coming back.
Also read: Chris Fall Resigns as Head of the US AI Safety Agency CAISI • Nvidia Pushes AI Agents Deeper Into 3D Design Tools at SIGGRAPH 2026 • Google Shares Rise After Report of a Chip Designed Around Gemini
