What AI Search Changes: When Page One No Longer Exists
The first page of ten blue links is disappearing: Google's AI Overviews passed 1.5 billion monthly users, and Baidu says most of its results are now AI-generated. Two timelines, three consequences for customer acquisition, and what to do about them.
Key takeaway
AI search replaces ranked links with direct answers: Google's AI Overviews passed 1.5 billion monthly users by May 2025, and Baidu said in November 2025 most results are AI-generated. Exposure shifts from clicks to citations — structured, well-sourced, direct-answer content now matters most.
Search a question on Google and an assembled AI answer now sits at the top of the page. Search the same question on Baidu and the first result is quite likely a rich-media answer rather than a list of links. Users no longer click through result after result; the answer is delivered to them. This is not a lab prototype — it is the deployed reality of 2025.
The overseas line: AI Overviews spreads, AI Mode launches
Google rebuilt search in two steps. Step one was AI Overviews, introduced at I/O in May 2024 to generate summary answers atop search results. By May 2025 it had passed 1.5 billion monthly users across 200+ countries and territories in 40+ languages, and Google said it lifted usage of relevant queries by more than 10% in markets such as the US and India — AI answers made people search more, not less.
Step two went further. On May 20, 2025, at I/O 2025, Google rolled out AI Mode across the US: a separate search tab built on a customized Gemini 2.5, using query fan-out to split one question into parallel sub-queries before composing the answer. The number worth remembering: queries in AI Mode run about two to three times longer than traditional searches. People have started throwing whole questions at the search box.
The domestic line: Baidu swapped out the results page
The Chinese internet moves to its own rhythm, but in the same direction. At Baidu World on November 13, 2025, founder Robin Li said the vast majority of Baidu's search results were already AI-generated, with rich-media coverage of the first result reaching 70% — the first thing a user sees is increasingly a direct answer in text, image or video rather than a link waiting to be clicked. The same event carried an easily missed figure: 625 vendors had connected to Baidu's search AI API through Baidu AI Cloud. Search capability itself is becoming an exported service.
Users didn't get lazy; they changed how they ask
Both lines point at the same behavioral shift: queries are longer and more conversational — from Shanghai bookkeeping price to how much would a twenty-person company in Shanghai pay for bookkeeping per year — and what people expect is a direct, complete, trustworthy answer rather than a pile of links. Doubled query length is not a detail. It means the search box is turning from a retrieval entry into a question entry — users have delegated the job of judging which page to trust to the AI.
Three consequences for customer acquisition
The first is traffic structure. When AI answers directly, a share of the visits that used to be click into the site to find the answer simply stops happening. Those still clicking have been pre-screened by an AI answer and arrive with clearer intent. Volume may fall; quality may not.
The second is the shape of exposure. Content value is no longer just about where you rank, but about whether AI cites you: if your definitions, data and methods are picked up and credited inside an AI answer, that is a new form of brand exposure — the user may never visit your site, yet remembers who supplied the answer.
The third is the content bar. When choosing what to cite, AI systems favor content that is clearly structured, factually sourced and answers the question directly; pages that read as patchwork with muddy claims do not even qualify for citation. The company website goes back to being material both humans and AI can read — exactly the direction we argued for in content assets.
What to do, and what not to do
The worthwhile work concentrates in three moves: write direct-answer content, one question with one clear answer, instead of circling for three paragraphs before the point; make it structured, with clean heading levels, distinct points and explicit data, so machines parse it easily; and cite sources for key facts, since verifiable content is easier for AI to trust. This emerging discipline has a name — GEO — unpacked fully in what GEO is and how to prepare.
What not to do is just as clear. Keyword stuffing and pseudo-original spinning do not merely fail in AI search — poor quality can exclude a page from the citation pool altogether. And services promising guaranteed AI inclusion or a locked slot in AI recommendations deserve the same skepticism as guaranteed first place on Google once did: citation mechanisms are opaque and constantly shifting, and nobody can guarantee them. For the broader content playbook, see what AI content marketing is.
A calm read: the trend is real, the pace is uneven
Separate the trend from the tempo. Directionally, search moving from links to answers looks all but irreversible. Tempo-wise, markets and industries diverge widely — the English-language ecosystem runs ahead, the Chinese ecosystem follows its own path, and B2B decision queries feel the shift differently from local-services queries. Our view: start building direct-answer, structured, well-sourced content as routine work now, watch the real traffic changes in your own industry, then size the bet. The trend is worth following; it is not worth a one-time all-in.