The U.S. Department of Justice won two consecutive antitrust victories in search and ad tech cases, and the court even considered forcing the divestiture of Chrome. Meanwhile, eMarketer predicts Google’s U.S. search ad market share will fall below 50% for the first time in 2025. But Google’s response is unexpected: it has chosen a more painful but more fundamental path — actively disrupting its own search model rather than defending the old order.
The Cracks in Advertising Hegemony
Meta is expected to surpass Google in 2026 to become the world’s largest digital advertiser, with Amazon and TikTok carving up advertising budgets. Google’s core search engine advertising revenue is facing its most severe structural challenge in two decades. But the real danger is not competitors stealing ad share — it’s that users are beginning to stop “searching.” Perplexity and OpenAI’s ChatGPT are replacing the “query-click-browse” behavioral pattern with a “ask-get answer” conversational model.
Defensive Moat: From Traffic Routing to Answers
Google’s counterstrategy is summarized as a “defensive moat.” On one hand, Google blocks AI rivals from scraping search index data through technical means — disabling the num=100 parameter is a typical case. On the other hand, Google is accelerating the embedding of generative AI into the core of search, launching AI Overviews and AI Mode, transforming the traditional “traffic routing model” into “agentic search” that directly provides answers.
Similarweb data shows that AI Overviews have expanded from appearing in 15% of searches to 43% in just one year, while AI Mode visits jumped from 126 million in June 2025 to 279 million in May 2026. What began as an AI layer on top of search has become an integral part of the search journey itself, with Google dropping users into AI Overviews, where they can continue research via AI Mode. This change has transformed Google from a gateway to the wider web into a destination in its own right.
The financial cost of this transition is enormous. Alphabet’s 2026 capital expenditure guidance was raised to $195-205 billion, with Q2 free cash flow at negative $5.9 billion. But the core business has not stalled: search and other revenue grew 17% year-over-year in Q2 to $63.27 billion. CEO Sundar Pichai’s statement is clear: “AI experiences are driving usage, and query volume has reached historic highs.”
Cloud and TPU: The New Vehicle for the AI Story
Google’s AI narrative is no longer told through the advertising business. Alphabet began recognizing revenue from TPU system sales in Q2 2026, sending self-developed chips into customer data centers rather than merely renting out compute time. Google Cloud revenue grew 82% year-over-year in Q2 to $24.77 billion, with backlog reaching $514 billion. Gemini Enterprise has been adopted by nearly 90% of Fortune 100 companies.
This means Google is replicating its vertical integration logic from search: from chips (TPU) to models (Gemini) to platform (Cloud) to applications (Search, YouTube) — full-stack self-control. TPU is the only self-developed chip solution capable of replacing Nvidia GPUs at scale, constituting Google’s structural advantage at the AI infrastructure level.
Talent and Organization: Painful Restructuring
Facing aggressive poaching by OpenAI, Meta, and Anthropic, Google DeepMind implemented a highly controversial non-compete strategy: paying full salary to place key talent on a one-year “paid leave” to block their immediate defection to rivals. About 20% of newly hired AI engineers in 2025 were returning former employees. Google Brain and DeepMind merged into a unified Google DeepMind with a unified AI infrastructure team.
Sundar Pichai admitted in an interview that in an enterprise of such scale, “what determines organizational agility is often not how perfect decisions are, but the speed of making them.” This painful “AI-first” restructuring reflects Google’s urgency to escape the “innovator’s dilemma.”
The Real Rules of AI Search: Content Depth
Google Search VP Liz Reid and Senior VP of Knowledge and Information Nick Fox gave nearly identical answers: AI search optimization works the same way as traditional search, but requires going one level deeper. Fox said at Google Marketing Live 2026: “Assuming AI will provide a high-level first-layer response, the best-performing content will be content that goes one or two layers deeper and is genuinely helpful there.”
Reid’s “three rules” also point to fundamentals: go deep into specific niches, publish content that AI cannot easily replace, and question GEO shortcut promises. Google’s official guidelines explicitly warn against “commodity content” — general summaries that repeat what others have published or that generative AI can easily produce — which offers little unique insight.
Search Behavior Itself Is Changing
Users are shifting from short keywords to two-, three-, and four-sentence natural language queries. This changes the information retrieval structure of search: more specific questions give smaller or specialized businesses a better chance to surface because their expertise can more closely match user needs. Gemini’s API processes more than 22 billion tokens per minute, up 37.5% from the previous quarter. More than 9 million developers use Google’s models monthly to build products.
But the data also shows Google’s usage still far exceeds ChatGPT. SimilarWeb data shows ChatGPT’s share of the search market is about 3%, despite rumors of it reaching 20%. Lily Ray continues to share these comparative data on LinkedIn, responding to claims that “no one uses Google anymore.”
Agentic AI and “Google Zero Traffic”
Alphabet is pushing Gemini toward “agentic AI” — no longer just conversation, but tool calling and task execution. Future search may no longer return blue links, but directly complete tasks through underlying agent tools. Pichai’s vision: when you search “plan a trip,” AI dispatches multiple agents in the background, generates a dedicated app interface and itinerary, and ultimately seamlessly converges to a unified user interface.
This has triggered a “Google zero traffic” panic in the publishing industry. Condé Nast’s CEO even told internal teams to treat Google search traffic “as zero” when making business plans. Pichai’s response was firm: Google remains committed to directing users to quality content, and the system prioritizes content from media based on users’ “paid subscription records,” but “low-quality clicks” will be naturally filtered.
Bottom Line: Google’s Real Bet
Google’s strategy is not to win the chatbot race — it is still catching up with ChatGPT on Gemini App’s 950 million monthly active users. The real bet is: ensuring that the “act of searching” itself is not replaced. If users no longer need to “search,” Google’s advertising empire loses its foundation. Therefore, Google would rather endure the traffic pain of “zero clicks” than allow users to flow to Perplexity or OpenAI.
Cloud and TPU’s 82% growth is the financial proof of this bet: AI investment is generating quantifiable revenue. But the scale of capital expenditure and the turn to negative free cash flow mean the margin for error in this transition is narrowing. Pichai’s cautious posture at Stanford’s graduation ceremony — proactively warning of AI bubble risk — may be the essence of Google’s strategy: using the certainty of infrastructure to hedge the uncertainty of the application layer.
