The PodKnowledge|← All Episodes
Based on Lenny's Podcast data
Lenny's Knowledge Sketch

Inside Google's
AI Search Turnaround

Robby Stein
VP of Product, Google Search
OCT 10 2025
The Context

Search Isn't Dying.
AI Is Expanding What We Ask

SEARCH AI INTEGRATION
"The core Google search isn't really changing, in my opinion. AI is expansionary. There's actually just more and more questions being asked and curiosity that can be fulfilled now with AI."
  • Three big components: AI Overviews, multimodal (Lens), and AI Mode
  • AI Mode: a full-screen way to ask Google anything, in natural language
  • Google Lens is growing 70% year-over-year in visual searches
  • 50 billion products in the Google shopping graph, updated 2 billion times an hour
Framework

Google's Search AI Strategy

AI SYNTHESISWEB LINKSAI MODE RESULTS
70%
year-over-year growth in Google Lens visual searches
50B
products in the Google shopping graph
  • AI Overviews: the fast AI at the top of the page
  • Multimodal: Google Lens, take a picture and ask about what you see
  • AI Mode: a full-screen experience with back-and-forth and follow-ups
  • Query fan-out: the model uses Google search as a tool, running dozens of background queries
The Robby framingAI Mode is designed and specially created for information. Planning a trip, buying something, working through a research question. Less focused on creativity or productivity.
Behind the Scenes

How AI Mode Went From Idea
to Ship in About a Year

  • The signal: people were literally typing "AI" at the end of their queries, trying to get the AI to do the thing
  • The start: five to ten people, a blank screen, "what if you could ask Google anything?"
  • Trusted testers: about 500 people externally, treated like a startup, friends and family telling us when it sucked
  • Then Labs: anyone could turn it on, real query data used to tune it, then a US launch and expansion
Query fan-out

When the AI constructs a response, it uses Google search as a tool. It appends dozens of other queries and searches in the background, then pairs each search with content.

Built for informational needs

How does it find information? How does it know if information is right? How does it check its work? These are all things we built into the model.

Playbook

Show Up in AI Answers

  • Do you satisfy the user intent — what they're trying to get?
  • Do you have sources? Do you cite your information?
  • Is it original, or is it repeating things that have been repeated 500 times?
  • Google's human rater guidelines still largely apply — those core signals are still valid and extremely useful
The creator moveIf I were a creator, I would be thinking, what kind of content is someone using AI for? And then how could my content be the best for that given set of needs?
Contrarian

What Robby Pushes Back On

Google is dead. Nobody wants to sit through search results.INSTEAD →The core Google search isn't really changing. AI is expansionary — more questions are being asked, not fewer.
You still need clever prompt incantations to get AI to do what you want.INSTEAD →Increasingly you can just use language. You don't need to do a lot of this heavy-duty fine-tuning.
Lean, scrappy teams are always the right answer.INSTEAD →There's this cult of lean, scrappy. People hold on to small teams too long and the product never gets good enough — it dies on the vine.
Differentiate the design to stand out.INSTEAD →Design for clarity instead of cleverness. If something's a standard and people understand it, lean into it.
Based on Robby Stein's episode on Lenny's Podcast. All ideas on this page are from the episode.Watch on YouTubeFollow @rmstein on X
Every sketch here began as a transcript.

Send us yours and we’ll draw it. $1 a transcript.

Sketch mine →
𝕏︎ X / Twitterin LinkedIn📸 Instagram🔗 Copy link
0:00