Inside Google's AI turnaround: AI Mode, AI Overviews, and vision for AI-powered search | Robby Stein

Key Concepts

• Gemini: Google's advanced AI model, recently topping app store charts.
• AI Overviews: Generative AI summaries appearing at the top of Google search results.
• AI Mode: An end-to-end, conversational, and multimodal AI-powered search experience (google.com/ai) designed for complex informational tasks.
• Google Lens: A multimodal AI product enabling visual search and understanding.
• Query Fanout: The process where Google's AI performs numerous background searches to construct a comprehensive response.
• Embodying Relentless Improvement: A product philosophy combining continuous, focused effort towards positive productivity with an unending drive to make things better.
• Jobs to be Done (JTBD): A framework for understanding the underlying problems or goals users "hire" a product to solve.
• Clarity over Cleverness: A design principle emphasizing straightforward, intuitive interfaces over overly innovative or confusing ones.
• S-curves: A model illustrating product growth phases: slow initial adoption, rapid growth, and eventual maturity/plateau.
• Search Live: A new conversational, voice-activated AI experience within the Google app.

Google's AI Transformation and Momentum

The speaker notes a significant internal shift at Google, characterized by an "incredible sense of focus and urgency" to deliver great products quickly. This renewed drive has led to Google Gemini hitting the number one spot in the App Store, a development few anticipated. Google's enduring mission to make information "universally accessible" is now more achievable than ever with AI. The current moment is seen as a "tipping point" where AI models can truly deliver for consumers.

Contrary to predictions that "Google is dead" due to the rise of chatbots like ChatGPT and Perplexity, the core Google Search experience is not changing but rather expanding. AI is "expansionary," leading to more questions being asked and curiosity being fulfilled. People use search for a "ridiculously wide set of things," from phone numbers and prices to directions and tax information, a vastness often "underappreciated."

Robbie Stein's Role and Background

Robbie Stein, the guest, is the VP of Product for Google Search, overseeing the entire search experience, including new AI Overviews, AI Mode, multimodal AI experiences like Google Lens, and the ranking algorithm. He is described as being at the forefront of one of Google's biggest shifts.

Prior to his current role, Stein was Head of Product at Instagram, where he led the launch of Instagram Stories, Reels, and Close Friends, contributing to Instagram's growth to half a billion daily active users. He was also on the founding team of Artifact and started two of his own companies (including Stamped, acquired by Yahoo). His career demonstrates a rare level of impact on global consumer products.

Google's AI Products: AI Mode, AI Overviews, and Google Lens

Google's AI strategy for search revolves around three main components:

1. AI Overviews: These provide quick, generative AI responses at the top of search results for natural language questions. They are "growing very, very quickly."
2. Multimodal AI (Google Lens): This enables visual search, allowing users to take pictures and ask questions. Google Lens is one of Google's fastest-growing products, with a 70% year-over-year increase in visual searches, already at a "massive scale" of billions. Examples include identifying shoes for purchase, getting homework help by photographing a question, or receiving book recommendations from a bookshelf photo.
3. AI Mode (google.com/ai): This is an "end-to-end frontier search experience" built on state-of-the-art models, allowing users to "ask anything of Google search" in a conversational, multi-turn manner. It leverages Google's vast information ecosystem, including:
   • 50 billion products in the Google Shopping Graph, updated 2 billion times an hour with live prices.
   • 250 million places in Google Maps.
   • Extensive finance information and the entire context of the web.

AI Mode is increasingly integrated into core experiences, allowing users to ask follow-up questions directly from AI Overviews or Lens results. The vision is a "consistent, simple product experience" where users don't need to think about where to ask a question.

The Evolution of Search and AEO/GEO

The speaker notes a return to the "Ask Jeeves" paradigm, where users can ask natural language questions and receive comprehensive answers, moving beyond keyword-centric searches. Users are now asking "real long hard complex questions" like "what's a great place for a date night... I already went to these four restaurants. I'm looking for outdoor dining and my friend has this allergy."

Regarding AEO/GEO (AI Engine Optimization), the AI constructs responses by performing "query fanout," where the model uses Google Search as a tool to execute dozens of background queries. It makes requests to Google's data backend for real-time information. The AI then pairs these searches with content. For content creators, the advice is to focus on creating "extremely helpful" web pages that satisfy user intent, provide sources, cite information, and offer originality, aligning with Google's long-standing human rater guidelines. Creators should consider "what kind of content is someone using AI for" (e.g., advice, how-to, complex needs) and tailor their content accordingly.

Google's AI is unique in its "unique access to Google Search signals," including spam detection and authority assessment, and its ability to link directly to authoritative sources. This differentiates it from other chatbots.

Lessons from Building AI Products

Robbie Stein highlights a recent, surprising trend in AI product development: the increasing "humanlike" nature of the interface. It's becoming easier to communicate with and steer AI using natural language, reducing the need for specific "incantations" or "hacks." Models are now "encoded to be able to say okay I'm going to like use more reasoning or thinking budget for that kind of a question or I'm going to use tools or code to code code use code execution." This "democratization of accessing these models" means less heavy fine-tuning is required for sophisticated outcomes.

Product Philosophy: Embodying Relentless Improvement

Robbie Stein's core product philosophy is "embodying relentless improvement," which he defines as:

1. Relentlessness: "Complete effort that is always exerted in a direction of positive productivity."
2. Making things better: "You have to always make things better. You're never content."

This philosophy stems from a deep "dissatisfaction" with the status quo, a drive to improve the world. He references Tony Fadell's "Think Younger" concept, which encourages questioning everyday frustrations (like fruit stickers) instead of habituating to them. This constant critical self-assessment and desire for improvement lead to a "compounding effect" that eventually hits a "tipping point" where products become genuinely useful and loved.

Metrics are not separate from this philosophy but are essential guides. They help product builders know if they are "on the right track" (e.g., J-curve for retention) and identify "root causes" when metrics drop.

Case Study: Instagram Stories and Close Friends

The launch of Instagram Stories was "quite controversial" as it adapted a format popularized by Snapchat. Stein argues that "not every great thing is going to be invented by you" and that denying users a better product by not adopting effective formats is detrimental. Instagram made Stories "its own" by adding unique creative tools, allowing photo uploads (unlike Snapchat at the time), and introducing features like pausing stories. This approach led to Stories feeling "Instagram" and "completed the product."

The Close Friends feature, which allows users to share private stories with a select group, initially "totally failed." Early versions were confusing, mistranslated "favorites" (leading to only one person on the list), and didn't clearly communicate its private nature. Through "analytical rigor" and "deeply understanding people," the team realized the "job to be done" was an "emotional job" of connecting with friends and sharing vulnerable moments. They found success when users added 20-30 people to their list, leading to more engagement. The solution involved:

• Simplifying it to be stories-only.
• Changing the name to "Close Friends."
• Building a list builder that recommended people.
• Designing for clarity: Placing the green ring outside the story bubble in the tray, making its private nature immediately obvious. This iterative process took "two or three years" but transformed a flop into a highly successful feature.

Strategic Resource Allocation and Growth

Stein emphasizes that product development involves a balance between optimizing existing features and making big bets on new ones. He uses the concept of S-curves to explain that as products or features mature, they reach "diminishing marginal returns." At these points, it's crucial to identify "next growth drivers" and invest in "first principled" new ideas. When a new feature creates a "new little growth engine," resources are then allocated to optimize it. Metrics serve as a "guide" to ensure impact and avoid "congratulating ourselves" without real user benefit.

The AI Mode Development Journey

The development of AI Mode began about a year ago, stemming from observations that users were trying to ask harder questions in AI Overviews and even appending "AI" to their queries. A small team of 5-10 people was formed to explore a "blank screen" experience for asking anything. An early, qualitative "moment of brilliance" (e.g., finding useful park information with maps and walkability) provided conviction.

The team then moved to a "trusted tester group" of ~500 external users (including friends and family for honest feedback), followed by a launch in "Labs" for broader user testing and data collection. This iterative process, driven by user feedback and data, led to its public launch at Google I/O in the US, with ongoing global expansion. Stein attributes this rapid development to a strong sense of "urgency" and the belief that the next year of product development will define how people use products for years to come.

Distinguishing Google's AI from Competitors

AI Mode is specifically "designed and specially created for information." It excels at tasks like planning trips, buying products, or research projects where users need "effortless information" and context, along with links to "authoritative sources." It is "less focused on things like creativity, productivity, like upload a spreadsheet and like output graphs for me." This positions it as an informational AI, distinct from more general-purpose chatbots.

Core Product Principles for Success

Robbie Stein outlines three (plus one) core product principles:

1. Deeply Understand People: Focus on the "jobs to be done" and "causation" – why someone "hires" a product. This involves "interrogation-style" interviews to uncover the "big hire" moment.
2. Analytical Rigor and Understanding Problems: Use data to dissect why metrics are dropping, perform "root cause analysis," and understand specific user pain points (e.g., the Close Friends mistranslation issue).
3. Designing for Clarity Instead of Cleverness: Prioritize intuitive design. If something is a standard, lean into it rather than reinventing it. He cites Don Norman's "Design of Everyday Things" and the example of confusing door handles. The green ring for Close Friends and the name "AI Mode" are examples of clarity.
4. Be Humble: Constantly question yourself, listen to users, and be open to being wrong.

Hot Take: The Need for Resources in Breakthroughs

Stein challenges the "cult of lean, scrappy, fast" teams, arguing that while internal conviction can start small, building "a product that works for a lot of people that is based on a technological breakthrough" often requires significant investment. He observes that teams sometimes "give up too early or underinvest," leading to products dying on the vine. The heuristic for scaling up is when "internal conviction" is high, there's "external validation," and the team is ready to "invest enough to make the best version of it" for launch.

AI Corner: Personal and Future Applications

Stein is particularly excited about AI's impact on "multimodal visual and inspirational needs." He believes AI will be "liberated to help in every possible modality," moving beyond text. He envisions future AI capabilities for inspiration, such as generating image boards for office designs (e.g., "mid-century modern beautiful office design with dark themes") and allowing multi-turn conversations with visual responses.

He also shares a personal application: his young children are using Search Live (a new conversational, voice-on AI experience in the Google app) to ask questions about animals, history, and school topics. This natural, voice-based interaction is helping them become "much more AI native."

Lightning Round Insights

• Recommended Books: Clayton Christensen's Competing Against Luck, Don Norman's Design of Everyday Things, and David Koepp's Aurora (fiction). He also mentioned reading Andy Weir's Project Hail Mary.
• Recent Media Enjoyed: The Bear, Dune, and Top Gun: Maverick.
• Favorite Product Discovery (non-digital): The "Purple Pillow," praised for its honeycomb polymer technology that provides support and prevents heat buildup.
• Life Motto: "Be curious," emphasizing the importance of wanting to know why everything is the way it is and chasing down knowledge.
• Startup Story (Stamped): As a 25-year-old co-founder, he cold-emailed Justin Bieber's manager, Scooter Braun, claiming to be in LA "tomorrow" (which wasn't true). This led to an immediate flight to LA, a meeting, and eventually getting Justin Bieber to use and promote the app, demonstrating the power of "intense urgency."

Conclusion and Key Takeaways

Robbie Stein's insights reveal a Google undergoing a rapid, focused AI transformation, driven by a deep commitment to its mission of universal information accessibility. His career underscores the critical role of "relentless improvement," a philosophy rooted in dissatisfaction with the status quo and a continuous drive to make things better. Success in product development, whether for new ventures or established giants, hinges on profoundly understanding user needs (jobs to be done), employing analytical rigor to solve problems, designing for clarity, and maintaining humility. The current AI era is seen as a pivotal moment, demanding strategic investment and a willingness to adapt and integrate new formats to meet evolving user expectations, ultimately expanding the utility of core products like Google Search.