SEO isn't dead. It's just that AI is running the show now
What Google's VP of Search said that should make us marketers uncomfortable.
About a year ago, I called AI search “the new marketing battleground.” I was right about the battlefield, just wrong about the timeline.
When I wrote that piece in June 2025, AI Overviews were still that weird little box at the top of Google that people kind of trusted… but not completely. GEO was a concept marketers had bookmarked, nodded at, and then done nothing about.
I recently watched Robby Stein — Google’s VP of Product for Search — on the Superhuman AI podcast, and it really got me thinking about where search is going. So this is my attempt to update the original piece with what Google has actually shipped, and what Stein let slip about where it’s heading.
What I got right, and didn’t
The shift from keyword optimisation to answer engine optimisation is real. Zero-click is real. Brands that structure content for machine readability are winning. GEO is now something SEO agencies actually pitch with a straight face.
What I underestimated? The sheer speed of Google’s moves. When I wrote this in mid-2025, AI Overviews were the main event. By November, Google had launched Gemini 3 Pro. By December, AI Mode was delivering faster, smarter answers — tackling complex questions, comparisons, and planning with stronger reasoning Search Engine Land globally. By January 2026, Personal Intelligence for AI Mode went public, connecting Gmail and Google Photos to deliver context-aware search responses.
The Robby Stein reframe: search isn’t shrinking, it’s expanding
The most counterintuitive thing Stein said is that AI isn’t cannibalising search. It’s making people search more.
Google is seeing more searches than ever as people ask harder, more conversational, and more visual questions powered by AI. AI hasn’t really changed the foundational needs: there’s just more curiosity being fulfilled now that AI makes it possible.
This is a meaningful reframe for marketers. We’ve been operating from a scarcity mindset — fewer clicks, less traffic, zero-click apocalypse — when the actual picture is closer to: the pie is getting bigger, but the slices are being cut differently.
What’s changing is the kind of query. Simple lookups still go straight to 10 blue links. But Stein described how people are increasingly asking harder, longer questions in natural language — and once you have a really specific question, AI Mode triggers with an overview and a way to dive into a back-and-forth conversation Limitless Podcast. You can now have 10, 20 turns with Google before landing on an answer.
For B2B marketers, this matters. Complex buying decisions — “which enterprise SaaS solution fits a hybrid team across Singapore, Manila, and Jakarta” — are exactly the kind of multi-variable queries where AI Mode thrives. Your content either shows up in that reasoning chain or it doesn’t.
Reasoning-aware search: a new approach to AI search
Here’s something I didn’t fully appreciate in my original piece: AI Mode doesn’t just retrieve information. It reasons. Stein described how Gemini-powered models issue their own internal Google searches behind the scenes — multiple simultaneous queries — to construct a single response.
And with Gemini 3 in the picture, the leap is significant. Gemini 3 in AI Mode can dynamically create the ideal visual layout for responses on the fly — featuring interactive tools and simulations — tailored to the specific query. When the model detects that an interactive tool will help, it uses its generative capabilities to code a custom simulation in real-time and adds it into the response.
Ask about physics? You get a live simulation. Ask about a financial comparison? You might get an interactive graph, not a table.
On the model side: Gemini 3 Flash is the default free model powering AI Mode globally, while users in the US can toggle “Thinking with 3 Pro” for complex, multi-step reasoning tasks — at slower speeds but deeper analysis. ALM Corp Think of it as economy versus business class reasoning, now built into the search bar.
Agentic search: “Done, I’ve booked it for you”
This is the part that should make every marketer pause.
In the podcast, Stein used a restaurant analogy that stuck with me. You’re looking for a date night spot in San Francisco. You get your AI-assisted shortlist. But then — Google has now implemented a live way for an agent to browse OpenTable and the web for reservations, then bring back not just great restaurants with reasoning and analysis, but also the specific times tables are available.
This is what agentic search actually means. Google is moving beyond providing links to executing complex, multi-step tasks by deeply understanding personal context. The “consideration phase” of buying — which is where B2B marketers spend most of their budget — is now being navigated by an AI agent on behalf of a human.
The implication for demand gen and content marketing is uncomfortable but worth sitting with: if an AI agent is doing the shortlisting, you need to be in the AI’s knowledge base, not just on page one of search results. Different problem. Same urgency.
Personalised search and the privacy question
The newest development, and the one with the most long-term consequences, is Personal Intelligence for AI Mode, which connects users’ Gmail accounts and Google Photos to let search reference personal emails, travel bookings, photo memories, and transaction histories when generating responses.
But the architecture it represents is the trajectory: search that knows your context, not just the world’s.
For B2B marketers in APAC, the honest answer is: this isn’t your immediate problem. But watch how it plays out. The brands that win in a personalised-search world are the ones already building trust signals into their content — being cited by AI as a credible source, not just ranking for keywords.
What this means for you now
My original piece was mostly about the shift from SEO to GEO. That still stands. But I’d add three updates:
1. Think about the reasoning chain, not just the citation. AI Mode constructs answers by reasoning across multiple sources. Your content needs to be usable in an argument, not just discoverable. That means being specific, factual, structured, and opinionated — not fluffy.
2. Agentic search is already here for transactional queries. If you’re in hospitality, travel, financial services, or retail, your booking/conversion funnel is being touched by AI agents right now. Figure out what that looks like for your category before your competitors do.
3. Visual search is growing faster than most APAC marketers realise. Stein mentioned a 70% year-over-year increase in visual searching on Google. Lens, circle-to-search, screenshot queries — this is an APAC behaviour pattern (anyone who’s watched how people shop in Southeast Asia on their phones knows this). If your visual assets aren’t optimised for machine readability, you’re invisible to this mode entirely.
The AI;DR
Elsewhere in the AIverse
Gemini can now create files on Microsoft 365 and Workspace. No more copy-pasting AI output into a doc like it’s 2022. Gemini now generates Docs, Sheets, Slides, Word, Excel, PDFs, and more directly in chat. For marketers churning out reports, briefs, and decks, this is a legitimate time-saver.
Twitter’s former CEO is building the search layer for AI agents — and investors are paying attention. Parallel Web Systems, founded by Parag Agrawal, just hit a $2B valuation after raising $100M — nearly tripling its valuation in five months. The startup builds internet and research APIs specifically designed for AI agents to search, extract, and monitor the web.
Vibe coding with Lovable just went mobile. Lovable — the no-code app builder that lets you spin up websites and apps from a text prompt — is now on iOS and Android. Kick off a build from your phone, let the agent run, get a notification when it’s ready.