Google Cloud AI Director Says SEO Must Adapt for Autonomous Agents
Addy Osmani says SEO must now serve agents that fetch and act on content, not people who click through pages. That turns clean structure and early answers into billable work.
SEO is shifting from a page-ranking contest to a machine-reading contest, and Addy Osmani is putting Google’s AI strategy behind that warning. The Google Cloud AI director said teams need to optimize for autonomous agents that fetch, parse, and act on content without the patience of a human browser, and he framed that work as agentic engine optimization, or AEO, not answer engine optimization.
Osmani’s background gives the message weight. He is a director at Google Cloud AI focused on Gemini, Vertex AI and the Agent Development Kit, after nearly 14 years leading Chrome’s Developer Experience. His current role bridges Google DeepMind, engineering, product and developer relations, which puts him close to the place where AI systems and web content now collide.
The practical advice is blunt: put the answer early, ideally within the first 500 tokens, because large pages can run past a model’s context window. When that happens, content can be truncated, skipped or misread. That makes semantic depth and retrievable structure central to visibility in AI search. It also changes what counts as good content architecture. Lead-ins have to be shorter, headings clearer and pages easier for machines to fragment and reuse without losing meaning.
For agencies, that opens a clear service line. An agent-readiness audit can map where key facts are buried too late in the copy, while structured-data cleanup can make products, services and FAQs easier for machines to parse. Task-completion content can be rewritten so an agent can pull a next step, a policy detail or a product answer without wading through marketing language. API and feed accessibility, clean HTML or markdown, robots.txt checks and firewall reviews become part of the scope. UX fixes matter too, because pages that hide core content from users or bots are already at a disadvantage in an agentic web where discovery, decisioning and transactions are increasingly handled by AI systems.
The larger shift is that AI SEO has become a planning category of its own. Search teams are already organizing around AI SEO, generative engine optimization and agentic AI in SEO, and Osmani’s framing gives agencies a practical brief: build content for consumption, not just ranking. In the agentic search era, the pages that win will be the ones machines can understand quickly and use without friction.
Know something we missed? Have a correction or additional information?
Submit a Tip
