Inside Google's AI turnaround: The rise of AI Mode, strategy behind AI Overviews, and their vision for AI-powered search | Robby Stein (VP of Product, Google Search)

Has Google finally found its second wind?

What if the search giant that once seemed sleepy has been quietly reinventing itself around the one thing it always owned: information? I was surprised by how personal that change felt when Robbie Stein described the mix of urgency, engineering and product craft powering Google’s recent momentum.

Not a single switch, but a tipping point

There’s no secret button that flipped. Rather, Stein portrays a compounding shift: close collaboration between product teams and DeepMind, relentless monthly improvements, and a sudden moment where models started delivering repeatedly for everyday needs. That narrative explains why Gemini climbed the App Store charts and why people are suddenly rediscovering Google.

AI that expands questions, not replaces search

Here’s what stood out: AI at Google is positioned as expansionary. Instead of cannibalizing queries, it surfaces more kinds of questions—visual, conversational, complex calculations—and helps people ask them in natural language. Google Lens is growing fast, and multimodal interactions are no longer a novelty but a core path to answers.

AI mode: a new front door to knowledge

AI mode isn’t just a flashy demo. It’s designed to be an end-to-end conversational layer that taps into Google’s living knowledge graphs: billions of shopping entries, live prices, maps and other real-time data. I liked how Stein emphasized that AI mode combines parametric model reasoning with active searching—what he calls "query fan out"—so answers can be both generative and verifiable.

Honestly, that hybrid approach feels like the most pragmatic route forward: leverage powerful reasoning but check it against the web and authoritative signals. That’s credible, and it changes what product teams should optimize for.

What makes AI mode distinct from other chatbots

• Designed specifically for informational tasks and search intent.
• Built with live access to Google signals and factual verification paths.
• Multimodal integration—text, voice, and visual search—so users can move naturally between modes.

Product craft matters more than ever

Stein’s central mantra—"embody relentless improvement"—is infectiously pragmatic. He tells a small, revealing story about asking his wife one-word impressions and getting back "dissatisfied," which he reframed as a fuel for obsessively making things better. That attitude carries through much of his career: Instagram Stories, Reels, Close Friends, Artifact, and now the Google Search frontiers.

Two things surprised me. First, major product wins rarely come from single brilliant strokes. They are long arcs of iteration, instrumentation, and ruthless user observation. Second, clarity beats cleverness. Names like AI Mode, or design affordances like the green ring for Close Friends, aren’t trivial—they reduce cognitive load and encourage adoption.

When incumbents must move fast

There’s a neat tension Stein navigates: startups race to capture early distribution while incumbents must adapt quickly or risk habit shift. He sees search as resilient because it serves such a broad set of needs, but also vulnerable when users change how they frame questions. The winning move: add complementary formats that feel native to your product rather than awkward add-ons.

Lessons from product launches—Stories to Close Friends

I appreciated a concrete case study: Close Friends. It nearly failed because of naming, UX choices, and cultural translation. They iterated for years—changing naming, list builders, and visual signals—and turned a flop into one of the platform’s most loved features. That arc is a reminder that metrics reveal problems, but user interviews reveal root causes.

Stein’s advice for creators aiming at AI-driven surfaces is equally practical: think about the job people are "hiring" your content or product to do. If your content answers deeper, advice-driven queries or emotional jobs, architect it differently than commodity pages. The fundamentals of good content still apply: original sourcing, clarity, and intent alignment.

Team sizing, curiosity and building momentum

Contrary to the lean-is-always-best narrative, Stein argues that some breakthroughs require real investment. Small teams can prove concepts, but hitting broad, high-quality scale often needs more resources. The heuristic: keep teams tight while you establish conviction; once you see external signals—real retention and compelling use—scale the team to ship the product that users actually need.

He returns again and again to curiosity as an operating principle. Be skeptical of the status quo. Ask why a flow exists. Chase the original sources. He mixes old-school reading with AI-assisted discovery, which felt like a balanced learning approach rather than a single-tool faith.

What I’m left thinking about

Google’s current play is part engineering sprint, part consumer product craft. The lesson is simple and maddeningly hard: invest in relentless iteration, instrument carefully, and design for clarity. If that sounds old-fashioned, that’s because the fundamentals of product still win even as interfaces become more human-like.

What if search stops being a list of links and becomes a living conversation that remembers context across modalities? That future will test how we value provenance, nuance, and curiosity—both as creators and consumers. It’s exciting, and oddly familiar: the tools change, but the work of asking better questions never does.