LLenny's Podcast
← All episodes
Robby Stein (VP of Product, Google Search)10 October 2025

Inside Google's AI turnaround: The rise of AI Mode, strategy behind AI Overviews, and their vision for AI-powered search

6Frameworks
15Insights

Frameworks in this episode

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 1

Myth Buster08:00

Why Search Isn't Dying: AI Is Expansionary

Despite predictions that ChatGPT and perplexity would kill Google, Stein argues the core search experience isn't changing. People still come to Google for a vast range of foundational needs (phone numbers, prices, directions), and AI hasn't replaced those. Instead, AI drives an expansion: more questions get asked and more curiosity gets satisfied, which is where the growth comes from.

  • Core Google search behavior and foundational needs are not being replaced by AI
  • AI is 'expansionary' — it creates more questions and more curiosity, not less search
  • The vastness of what people search for is underappreciated
  • Growth comes from the expansion moment, not from cannibalizing existing search

AI is expansionary and so there's actually just more and more questions being asked and curiosity that can be fulfilled now with AI

Robby Stein · 08:30

the core Google search isn't really changing in my opinion we're not seeing that but you're getting this expansion moment

Robby Stein · 09:00
#google-search#ai#growth#product-strategy

Hot Take· 2

Hot Take34:00

Formats Become Primitives — And You're Robbing Users If You Skip Them

Stein pushes back on the idea that copying a format is theft. Facebook may have created the modern feed, but now there's a feed for every product — DoorDash, LinkedIn, and more. Once something becomes a core primitive, refusing to adopt it just denies your users a better product for their specific use case.

  • Facebook created the modern feed, but feeds now exist across every product
  • Core primitives and formats get adopted quickly across the industry
  • Refusing to adopt a proven format robs your users of a better product
  • Each product uses these formats differently for its own users

at the end of the day you're kind of just robbing your user base of the opportunity to have a better product

Robby Stein · 34:30
#product-strategy#feeds#instagram#hot-take
Hot Take1:03:00

Hot Take: The 'Cult Of Lean' Makes Teams Give Up Too Early

Against the Twitter narrative of tiny scrappy teams, Stein argues that big breakthroughs built on technological advances often need real investment. Teams frequently underinvest and give up too early, or hold onto small teams too long so the product never gets good enough and dies on the vine. He advises thinking from first principles about the group you actually need to ship a great version.

  • There's a 'cult of lean' — scrappy, fast, throw away your product quickly
  • Hard technical breakthroughs took hundreds of people and years (e.g. foundational models)
  • The bigger risk is holding small teams too long until the product dies on the vine
  • Decide the team from first principles, not by blindly staying two people until product-market fit

there's kind of this cult of lean, scrappy, fast, throw away your product quickly, keep moving.

Robby Stein · 1:03:00

I see the opposite more true where people hold on to small teams too long

Robby Stein · 1:04:30
#team-building#startups#hot-take#product-strategy

Explainer· 4

Explainer09:00

Google Lens: Point Your Camera At Anything

Stein describes Google Lens as one of the fastest-growing products, with visual searches up 70% year-over-year at billions of scale. He gives concrete examples of the multimodal use cases now possible: photographing your shoes to buy them, snapping a stuck homework question, or pointing at your bookshelf for reading recommendations.

  • Visual searches growing 70% year-over-year, already at billions of scale
  • Take a picture of your shoes and ask where to buy them
  • Photograph a homework question you're stuck on and get help
  • Point at your bookshelf and ask what books to get next

It's growing 70% um year-over-year increase in visual searches, which is already at like a massive scale.

Robby Stein · 09:00
#google-lens#multimodal#visual-search#ai
Explainer15:30

Query Fanout: How AI Search Actually Works Under The Hood

Stein explains that when Google's AI constructs a response it does something called 'query fanout' — the model uses Google search as a tool, appending dozens of background queries and pairing each with content. His advice for the AEO/GEO era: the old signals of genuinely helpful, sourced, original content still apply, and creators should think about what people actually use AI for.

  • The AI performs 'query fanout' — dozens of appended background searches, each paired with content
  • Google's human rater guidelines and core quality signals still apply
  • Best practices: satisfy user intent, cite sources, be original not repetitive
  • Creators should ask what content people use AI for and be the best answer for that need

it's actually trying to does something called query fanout where the model uses Google search as a tool

Robby Stein · 15:30

It'll add and append all these other queries like maybe dozens of queries and start start searching basically in the background

Robby Stein · 16:00
#ai-search#seo#geo#content-strategy
Explainer50:00

How AI Mode Is Positioned Against ChatGPT And Perplexity

Asked whether AI Mode is a direct ChatGPT competitor, Stein frames it as different: purpose-built for information, not a general chatbot. It's designed for the things people come to Google for — planning a trip, buying something, research — with links to verify against authoritative sources. It's deliberately less focused on creativity and productivity tasks like generating graphs from spreadsheets.

  • AI Mode is specially designed for information, not general creativity
  • Built for classic Google jobs: planning, buying, research questions
  • Provides links to verify and dig into authoritative sources
  • Deliberately not focused on productivity tasks like spreadsheet-to-graphs

AI mode is a way to ask search anything you want it's for it's designed and specially created for information

Robby Stein · 50:00

we are less focused on, you know, creativity, productivity, like upload a spreadsheet and like output graphs for me.

Robby Stein · 50:30
#ai-mode#positioning#chatgpt#google-search
Explainer1:08:00

A Visual AI Mode For Inspiration (Pinterest, Beware)

In the AI corner, Stein previews a visual version of AI Mode announced at IO and being rolled out. Instead of describing ideas in text, you can ask for a design vibe — like a mid-century modern office with dark themes — and get an inspirational image board you can refine multi-turn ('more coastal,' 'more creamy'). It targets inspirational needs (decor, shopping) rather than utilities like code and homework.

  • AI grew up in text; it's now being liberated into every modality
  • A visual AI Mode produces inspirational image boards from a described vibe
  • You can multi-turn on the images with natural language to refine the aesthetic
  • It's distinct from Nano Banana (an image editor) — this is AI image search/inspiration

what's a what's a mid-century modern beautiful office design with dark themes? it'll be able to produce this image board that's inspirational and you can…

Robby Stein · 1:08:30
#ai-mode#multimodal#inspiration#google

Story· 4

Story22:00

The One Word His Wife Used To Describe Him: 'Dissatisfied'

Backstage at an Instagram all-team icebreaker, Stein texted his wife asking for one word to describe him — she replied 'dissatisfied.' At first offended, he realized she meant it as a driving force: a deep desire to make the world better. He ties this to 'relentless improvement' — being the harshest critic of your own work until compounding effort tips a product into being genuinely useful.

  • His wife's one-word description of him was 'dissatisfied'
  • It meant not unhappiness but a drive to make the world better
  • Successful products shared a spirit of 'two more moves to make it better'
  • Be the harshest critic and most dissatisfied person about your own work

First thing that comes to your mind and she just wrote back dissatisfied.

Robby Stein · 22:30

One is just relentlessness like just complete effort that is always exerted in a direction of positive productivity.

Robby Stein · 22:00
#product-mindset#leadership#instagram#improvement
Story30:30

Instagram Stories: Not Every Great Idea Has To Be Yours

Reflecting on the controversial Stories launch that copied Snapchat, Stein argues you have to know when a format is an existential question and learn from the world. Stories succeeded because Instagram made the format its own — letting users upload high-res photos from their camera roll (Snapchat forbade uploads) and adding a finger-hold pause. Those small dissatisfaction-driven fixes made it feel native.

  • Not every great format will be invented by you — formats become primitives you can make your own
  • Snapchat didn't let you upload photos; Instagram did, treating stories as savable memories
  • Snapchat stories couldn't be paused; Instagram added finger-hold to pause
  • Small, native touches made Stories feel like it 'completed' Instagram

not every great thing is going to be invented by you

Robby Stein · 31:00

why don't you just let people upload their c their photo

Robby Stein · 33:00
#instagram#product-strategy#stories#snapchat
Story57:00

Instagram Close Friends: A 3-Year Flop Fixed By An Emotional Job

Close Friends took two to three years and initially failed completely — mistranslated as 'favorites' in many markets, so people added only one or two people and never got a reply. Stein reframes the real job: it's emotional, not utility. The product exists to connect you to friends, and if you don't get a DM back, it's broken. Fixes — a 20-30 person list builder, renaming, and moving the green ring outside the story — turned the flop into a hit.

  • Close Friends took 2-3 years and totally failed at first
  • Mistranslation as 'favorites' led people to add only 1-2 people, so no replies came
  • The real job is emotional — connection — not utility; success = getting a DM back
  • It worked when people added 20-30 people; a list builder, rename, and outer green ring fixed it

there are utility jobs and there are emotional jobs. People usually discount the emotional ones a lot.

Robby Stein · 59:30

the job that we're doing is actually connecting you to your friends. And if you don't get a DM back, it's broken.

Robby Stein · 59:00
#instagram#jobs-to-be-done#product-fixes#close-friends
Story1:17:30

How Booking Justin Bieber Taught Him 'Intense Urgency Wins'

At 25, building his startup Stamped, Stein and his co-founder cold-emailed Justin Bieber's manager Scooter Braun claiming they'd be in LA 'tomorrow' — when they weren't. Braun replied with a one-line invite to breakfast, so they went straight to the airport, flew out, and pitched. Bieber ended up stamping his favorite things on the app, drawing lots of new users. The lesson: be scrappy, be immediate, and let intense urgency beat overthinking.

  • They cold-emailed Scooter Braun claiming they'd be in LA tomorrow — they wouldn't be
  • A one-line reply invited them to breakfast, so they flew out immediately
  • Getting Bieber to 'stamp' his favorites drew lots of users to try the app
  • Lesson: intense urgency and scrappiness usually beat thinking about it for a long time

he he just like wrote back some oneline thing like meet me at this hotel like for breakfast

Robby Stein · 1:19:00

do it now be scrappy be immediate like intense urgency usually wins over uh thinking about it for a long time

Robby Stein · 1:20:00
#startups#hustle#growth#stamped

Tool· 2

Tool1:12:30

Search Live: Talk To Google Out Loud (Great For Kids)

Stein reveals Search Live launched out of labs that week — a conversational, voice-on way to talk to Google search live, accessible via a 'live' button on the Google app home screen. He finds it especially accessible for kids, who come home wanting to 'talk to Google' about animals or history, making them more AI-native than anything else he's tried.

  • Search Live lets you talk to Google search in a live, voice-on conversation
  • Access it via a 'live' button on the Google app home screen — a full-screen experience
  • Useful hands-free, e.g. while driving
  • His kids use it naturally to ask about animals and history, becoming more AI-native

you can talk to search in a live AI setting, which is conversational, voice, voice on

Robby Stein · 1:13:00
#search-live#voice-ai#google-app#kids
Tool1:16:00

Robby Stein's Non-Digital Product Pick: The Purple Pillow

In the lightning round, Stein recommends a non-digital product he's been evangelizing to coworkers — the Purple Pillow. He describes its honeycomb polymer technology that supports your head while micro-holes keep it from getting hot, and says his team now has a running 'pillow chat.'

  • Recommends the Purple Pillow, a non-digital product
  • Uses a honeycomb polymer that supports your head
  • Micro-holes keep the pillow from getting hot
  • He's been recommending it widely at work

I'm super into this new pillow that I got called Purple Pillow

Robby Stein · 1:16:00

it's got this like new technology of like this honeycomb polymer that's inside and so it like supports you and it has these little micro…

Robby Stein · 1:16:30
#product-recommendation#lightning-round#pillow

Takeaway· 2

Takeaway19:00

Steering AI Is Now Humanlike — Fine-Tuning Is Fading

Stein observes a shift in just the last months: you no longer need 'incantations,' hacks, or heavy post-training to get sophisticated behavior from a model. You can now write instructions in plain language — describe your APIs, schema, and when to use them — and the model decides to reason more, use tools, or run code. He believes this democratizes building powerful things without heavy-duty fine-tuning.

  • You used to need prompt hacks and 'act like a coach' incantations to steer AI
  • Sophisticated behavior once required post-training and updating model weights
  • Now plain-language instructions can make the model reason more or use tools/code
  • Heavy-duty fine-tuning is increasingly unnecessary for sophisticated outcomes

it's so obvious how humanlike the interface is becoming with how you can communicate and steer AI.

Robby Stein · 19:00

increasingly I don't think you need to do a lot of this heavy duty fine-tuning

Robby Stein · 20:30
#ai#prompting#fine-tuning#product-building
Takeaway54:30

Design For Clarity, Not Cleverness (Doors And Camera Icons)

One of Stein's core principles is designing for clarity instead of cleverness. Citing Don Norman's 'The Design of Everyday Things,' he notes that beautifully symmetric door handles fail to tell you whether to push or pull. The same trap shows up in product: teams design clever AI-looking camera icons when a plain camera icon communicates instantly and gets people to use the product.

  • Leaning into a recognized standard gives you leverage; reinventing it costs you
  • Don Norman's door example: clever symmetric handles hide push-vs-pull
  • Teams over-design icons when a plain, global icon communicates instantly
  • Be thoughtful about when to reinvent versus when to lean into a standard

you walk up to a door and based on how they're designed at times, people still don't know if you should pull or push that…

Robby Stein · 55:00

it's a camera just just put the camera in maybe you could add like a little thing to it

Robby Stein · 55:30
#design#product#ux#clarity