Sundar Pichai this week sat for his fifth consecutive post-I/O interview on The Verge's Decoder podcast, hosted by Nilay Patel, in a conversation recorded on May 26, 2026, just days after the Google I/O developer conference. The discussion covered the structural reorganisation Google undertook after ChatGPT, the mechanics of Gemini Spark and Antigravity, the deteriorating relationship between the company and web publishers, and Pichai's reluctant acknowledgment that at least one current Search experience is more opinionated than it should be.

The interview is one of the more candid public appearances Pichai has given in recent years, touching on questions that Google has historically deflected - including whether publishers are right to plan for zero search referral traffic and how close the industry is to AGI.

How Google reorganised itself after ChatGPT

Pichai described the period following ChatGPT's launch in late 2022 as one requiring a structural response, not just a product response. According to Pichai, the company had all the underlying technology but was not organised to move at the speed the moment demanded.

The first significant change was consolidating the company's two leading AI research organisations. Google Brain and DeepMind had both operated as world-class independent entities, but Pichai brought them together as Google DeepMind. He acknowledged the difficulty of that process, describing it as "like saying, go put Stanford and MIT together and create a department out of it." Alongside that consolidation, Amin Vahdat was appointed as SVP of AI Infrastructure to lead a centralised infrastructure team, and Koray Kavukcuoglu took on the role of Chief AI Architect - a new position designed to oversee how AI technology is applied across the whole of Google.

Search itself was restructured. Previously distributed across several leaders, the product was pulled under Elizabeth Reid, with Nick Fox holding responsibility for the broader area and Josh Woodward moving to lead the Labs product before shifting focus to Gemini. Pichai also described setting up weekly AI product reviews that he ran personally, requiring that anything being shipped to users went through that channel directly. The intent, he said, was to make the company capable of faster decisions, because decision velocity matters more than the content of any individual decision in most cases.

The reorganisation is a matter of public record, but Pichai's framing on Decoder places it in a specific sequence: he saw the Overton window shift - meaning the pace at which people were adopting AI technology moved faster than Google had modelled - and concluded the company needed to reorganise around a common infrastructure layer rather than continue building AI capabilities product by product.

That common layer is now the Gemini model family. Pichai described Personal Intelligence - which integrates a user's Gmail, Google Photos, YouTube history, and Search history into AI responses - as an example of a capability built once and applied across products. Google expanded Personal Intelligence to nearly 200 countries at I/O 2026, having previously restricted it to US beta subscribers.

Gemini Spark, Antigravity, and what agents actually mean

A significant portion of the interview focused on the agentic layer Google announced at I/O. Patel asked whether agents represent a stable paradigm or whether another shift is coming, and Pichai's answer was more grounded than typical conference framing. According to Pichai, the building blocks now in place - the ability to reason, use tools, and write code - constitute something close to a complete primitive set for what agents need to do.

Gemini Spark is positioned as a tab within the Gemini app rather than a standalone product, a background agent that continues operating even when a device is closed. As PPC Land reported after I/O 2026, Spark runs on Gemini 3.5 Flash and connects to Google Workspace - Gmail, Docs, Calendar, Slides - along with external services including Canva, OpenTable, and Instacart, initially in the United States only. Pichai described Spark as the consumer expression of capabilities that Google's own developers are already using internally.

Antigravity, the agentic coding platform, sits on the developer side of the same infrastructure. Pichai noted that a portion of Google's engineers have already moved beyond using AI tools to assist their work and are now directing teams of agents. The Antigravity engine, he said, is built into Gemini itself - the same harness powering developer tooling is the one serving consumers, just accessed through a different interface. PPC Land's analysis of the I/O 2026 agentic panel recorded Koray Kavukcuoglu describing the outcome as a virtuous cycle: faster tools lead to better models, which enable faster tools again.

Patel pushed Pichai on whether the Search box, Spark, and Canvas - which can generate custom software in response to planning tasks - should simply be one product. Pichai agreed they would converge, comparing the trajectory to how Notebooks started as a standalone innovation in NotebookLM and are now appearing consistently across Gemini. He used the same logic for agents: the user should not need to think about which surface a task runs on.

The I/O 2026 developer keynote introduced Antigravity 2.0 and a proposed open web standard called WebMCP, which aims to replace the fragile screenshot-and-DOM-parsing approach current browser agents use with direct tool calls. That proposal matters for marketing because it affects how agents interact with websites - and therefore how publishers and advertisers can expect to be discovered or bypassed.

The Search result Pichai called too opinionated

One of the more striking moments came when Patel showed Pichai a search result on his phone - a query for "best Chromebook" that Patel said he has been tracking for years. The result opened with an AI Overview giving a confident recommendation, followed by sponsored boxes, and then an organic Reddit result offering a different answer. Below that, a news publication offered a third answer.

Pichai looked at it and said the result was "probably more opinionated than it should be" for that particular query. He suggested there was scope for improvement, though he also floated the possibility that the result had been personalised specifically for Patel given his browsing patterns - a framing Patel pushed back on by noting his broader point about infinitely personalised results producing inconsistent experiences at scale.

The exchange is notable because it cuts to a structural tension that PPC Land has tracked since the early rollout of AI Overviews. AI-generated answers at the top of results pages are not neutral summaries - they carry an editorial posture, and that posture is not always consistent with the organic content sitting below. Ahrefs research published in April 2025 found a 34.5% reduction in organic clicks when AI Overviews are present in results, a figure that grew to 58% by February 2026 in a follow-up study. For marketers, that trajectory means the stakes attached to how Google's AI interprets and surfaces content have risen sharply.

Pichai did not dispute the direction of the data. He said Google measures user satisfaction through engagement signals - sessions, bounce rates, return visits - and that the product evolves based on those metrics over the long term. But he conceded the specific result Patel showed him illustrated a real gap between what the AI was expressing and what the query probably called for.

Google Zero is no longer theoretical

Patel introduced the term Google Zero - the idea that Google traffic to websites will eventually approach nothing as queries are answered directly within the search results page - and noted that Roger Lynch, CEO of Conde Nast, had told TBPN in a May 2026 interview that the company has instructed its teams to plan as though search traffic does not exist. Conde Nast's position, which PPC Land covered in detail, follows three consecutive years in which the publisher's traffic forecasts proved too optimistic. Each year, traffic declined more than the company's internal models had projected, leading Lynch to set a baseline of zero rather than continue building budgets around a declining variable.

According to Pichai, publishers understand their own businesses better than he does, and he is not in a position to tell an iconic publisher how to plan. He committed to reflecting high-quality content in Google's products, but did not dispute Lynch's underlying characterisation of the trend.

His broader response was to frame search traffic decline as part of a wider information ecosystem shift rather than a Google-specific phenomenon. Referral traffic to publishers now comes from a much wider set of sources than it did a decade ago, he argued, and bounce clicks - users clicking a result and immediately returning to Google - are declining, which he interpreted as higher engagement quality on the clicks that do happen.

That argument has been contested repeatedly in independent research. A study of UK businesses found an 86% collapse in organic traffic growth across 800 companies after Google deployed AI Overviews and AI Mode. PPC Land has also tracked how zero-click searches reached 69% of all Google queries by mid-2025, up from 56% when AI Overviews launched in May 2024. The gap between Google's quality-click framing and the volume-based reality publishers experience has not narrowed.

Patel also asked whether YouTube creators should have the same ability to opt out of AI training as publishers are debating in court. Pichai gave an indirect answer, noting that Google had offered opt-out capability through Google-Extended and that the company is in ongoing conversations with publishers. He acknowledged the company is engaged in litigation with publishers in the UK, where Google has described a proposed publisher arrangement as a "free writer charter" - a characterisation the News Media Association disputes.

Public anxiety is structural, not a marketing problem

Asked directly whether the gap between user satisfaction metrics and public polling on AI could be closed through better marketing, Pichai rejected the premise. According to Pichai, the anxiety people feel about AI is rational and goes deeper than product experience. People are hearing that jobs may be eliminated, energy prices may rise because of data center construction, and the pace of change is faster than human cognition is naturally equipped to process.

"I don't think humans are evolved for processing this much change," Pichai said in the Decoder interview. He described the rate of change over the last several years as "incredibly high" and said people are trying to understand what it means for their lives at an economic level, adding that it "really makes sense" why anxiety exists.

He pointed to SynthID - Google's tool for watermarking AI-generated content - and to a Ratepayer Pledge the company has signed alongside other major technology firms, as examples of the industry taking structural responsibility. But he stopped short of claiming that either tool resolves the underlying concern. The public, he said, needs to have a voice in how the most consequential technology of this era gets deployed in democratic societies.

Pichai also discussed AI slop - a term increasingly used to describe low-quality AI-generated content proliferating across the web - and said he feels it. He attributed its presence partly to competitive dynamics in an early phase of technology deployment, where speed to market creates incentives to publish before quality standards are established.

What Demis Hassabis meant by the foothills of the singularity

The I/O 2026 keynote closed with Demis Hassabis, CEO of Google DeepMind, telling the audience that "when we look back at this time, I think we will realize that we were standing in the foothills of the singularity." Patel asked Pichai to explain what that means practically.

According to Pichai, his conversations with Hassabis have produced an alignment on how both men define AGI: a system that can comprehensively perform the wide range of cognitive tasks a human can perform, at a comparable level. The "foothills" framing reflects a view that current systems are approaching but have not yet reached that threshold. Pichai said Google is working on a formal public statement of its AGI definition.

On timeline, Pichai declined to name a specific year. His argument was that the rate of progress makes the AGI threshold less important than what the systems can do as they approach it. "The way I would answer that question three years from now, whether you and I call it AGI or not doesn't matter because it'll be very, very powerful and we have to prepare for it," Pichai said. He noted that consensus exists among frontier labs - not just Google - that the technology is progressing faster than most outside observers realise. He placed the window in the range of three to five years as a rough industry conversation point, while emphasising the uncertainty around the edges of that range.

Whether LLMs can get there was Patel's next question. Pichai compared the transformer architecture - invented at Google and still the foundation of modern language models - to the von Neumann architecture underlying modern computing. The underlying design persists, but what runs on top of it has changed almost beyond recognition. He described the last three years of trajectory as remarkable and said he sees no reason to assume that trajectory will plateau, while acknowledging that fundamental breakthroughs in scientific reasoning - machines making genuinely novel discoveries without human prompting - have not yet arrived.

Pichai's April 2026 conversation on the Cheeky Pint podcast covered related ground, including a $175-185 billion capital expenditure plan for 2026 and the argument that Search will become an agent manager within a decade - a framing that maps directly onto what Gemini Spark and Antigravity represent at I/O 2026.

Why this matters for the marketing community

The Decoder interview surfaces several dynamics that the marketing industry is navigating in parallel. The admission that Search results can be more opinionated than warranted - even from the CEO's perspective - confirms what performance marketers have observed: the AI layer introduces editorial variance that paid and organic strategies have to account for differently than traditional ranking factors.

The Conde Nast zero-traffic posture is significant not as an isolated outlier but as a benchmark. If one of the largest and most resourced publishers in the world is building its business plan around the assumption that Google referral traffic is negligible, the signals available to smaller publishers and brand-owned content operations point in the same direction. PPC Land has tracked the structural decline since Google's early AI Overview rollouts, and the trajectory documented in third-party research has consistently been steeper than Google's own characterisations suggested.

Gemini Spark's integration with Workspace and external services through MCP also has direct implications for how agentic search intersects with advertising. The I/O 2026 search announcements include information agents that monitor the web continuously without user prompts - a model where the traditional search funnel, in which a user types a query and receives results, no longer applies to a portion of decisions. Campaigns optimised for query-triggered discovery will need rethinking as that portion grows.

Timeline

Summary

Who - Sundar Pichai, CEO of Google and Alphabet, in conversation with Nilay Patel, Editor-in-Chief of The Verge, on the Decoder podcast. Roger Lynch, CEO of Conde Nast, is referenced as a key third-party voice on publisher search traffic.

What - Pichai discussed Google's post-ChatGPT structural reorganisation, the mechanics of Gemini Spark and Antigravity as consumer and developer expressions of the same agentic infrastructure, the deteriorating referral traffic relationship between Google Search and web publishers, a specific Search result he described as more opinionated than warranted, public anxiety about AI as a structural rather than marketing problem, and AGI timelines in the range of three to five years as a rough industry consensus.

When - The interview was recorded on May 26, 2026, shortly after the Google I/O developer conference held on May 19, 2026. It is the fifth consecutive annual Decoder conversation between Patel and Pichai following Google I/O.

Where - The conversation was published on The Verge's YouTube channel at https://www.youtube.com/watch?v=ANV3tE5ywv0. The I/O conference it references took place at Google's campus in Mountain View, California.

Why - The interview matters because it is one of the few occasions on which Google's CEO has publicly conceded imperfections in current Search results, acknowledged that a major publisher's zero-traffic posture is a legitimate business response, and described the public's anxiety about AI as rational rather than a communications failure. Each of those positions has direct implications for the marketers, publishers, and developers whose businesses depend on how Google routes attention across the web.

Share this article
The link has been copied!