Google I/O 2026 became one of the most-watched developer keynotes in the company's history and, at the same time, one of the least applauded. The gap between those two facts is the story.
The annual conference ran in Mountain View, California, with the keynote streamed worldwide. According to PPC Land's coverage of the search changes announced at the event, the keynote took place on May 19, 2026, and the advertiser-facing Google Marketing Live followed one day later. The product list was long. The reception, by at least one measure, was thin.
On May 26, 2026, the technology podcast The WAN Show published a segment titled "Nobody Liked Google I/O 2026," in which hosts Linus and Luke walked through the announcements and the audience response to them. Their central observation concerned the keynote video itself. "I don't think I have ever seen a video with such a low interaction to view ratio," one host said of the official recording. He put a number on it: the share of viewers who pressed the like button worked out to "something like 0.05%." The video, he noted, had reached "6.8 million" views at the time he checked.
A separate figure circulated in the comments under the WAN Show video. One commenter, posting a tally dated May 25, recorded the official keynote at 17.6 million views against roughly 12,500 likes and 25,900 dislikes - a count in which negative votes outnumbered positive ones. Neither set of numbers can be independently verified here, and view counts and vote tallies shift continuously. What is consistent across both is direction: a very large audience, and a strikingly small share of it choosing to signal approval.
The reason the numbers matter is the content they attach to. Google used the keynote to push artificial intelligence into nearly every consumer product it operates, and the WAN Show hosts did not hide their unease about that. One host said the approach "is pretty scary and would also qualify for a bad news show," even as he framed the audience reaction as the more interesting development.
The announcements form a dense package. The new Gemini 3.5 family of models headlined the event. Gemini 3.5 Flash shipped immediately, and the hosts noted that Gemini 3.5 Pro was described as launching the following month. PPC Land's analysis of the search changes confirms the timing and adds the technical detail that Google replaced the previous default model inside AI Mode with Gemini 3.5 Flash as of May 19, across every country where AI Mode operates. The same coverage reports that AI Mode has surpassed one billion monthly users, and that queries in AI Mode run on average three times as long as conventional search queries.
The headline claim about the new model concerned speed rather than raw capability. The WAN Show hosts summarised the pitch as a model "faster and better at agentic tasks." PPC Land's reporting, drawing on commentary from SEO analyst Marie Haynes, records Sundar Pichai stating at the event that Gemini 3.5 Flash "tends to be about four times faster than most of the Frontier models" and, inside Google's Antigravity infrastructure, "it's 12 times faster than that." Haynes also relayed the benchmark assertion that 3.5 Flash "for most things is actually better than Gemini 3.1 Pro" - a Flash-tier model overtaking a previous-generation Pro model. That speed gain, PPC Land notes, is what makes persistent background agents technically feasible at scale.
The agentic theme ran through the rest of the lineup. Google announced Gemini Omni, which the WAN Show described as a new family of models generating video clips from combinations of text, photo, video and audio inputs. The hosts singled out a demonstration in which a single clip was re-rendered from twelve camera angles and then shifted from daytime to nighttime on command. One host, recalling the manual effort the same task required during his earlier career in video production, conceded the demo "was pretty cool."
Google also showed Gemini Spark, described by the hosts as "an AI agent that can command other AI agents and interact with Google Workspace apps." PPC Land's account of the search announcements describes Spark as a personalized assistant running on Gemini 3.5 Flash, integrated with Gmail, Docs, Calendar and Slides, and able to keep running even when a device is closed by using the Antigravity infrastructure. The WAN Show hosts, usually sceptical of Google's stage demos, found this one plausible. One said it was "the first time I've seen one of these Google demos and I've gone, okay, that does actually seem potentially feasible." The example that convinced him - an agent collecting party RSVP emails into a spreadsheet - was, in his words, "exactly the sort of tedious nonwork that some stupid AI agent could go do for me."
That qualified optimism was the exception. One agentic use case drew open discomfort. Google highlighted agents that monitor for product availability and concert ticket releases and then complete the purchase automatically. "I can't believe that we're just talking about this openly now, but ticket scalping," one host said, observing that automated buying would make limited-release items harder to obtain for anyone without an agent of their own.
The same monitoring capability appears in Search. PPC Land's coverage following the keynote, drawn from a panel session Google published on May 22, 2026, places this background-agent design at the centre of how Search is being rebuilt. Search agents can now watch continuously for changes such as stock movements or price drops on shopping listings, and Search can generate small applications and visualizations on the fly. The WAN Show hosts flagged a trust problem with the last feature. Shown a generated visualization explaining gravity waves, one host said it "looks really cool" but added that verifying its accuracy would require checking an independent source that a person already trusts - which removes much of the time-saving the feature promises.
Beyond software, Google showed an updated version of its smart glasses, developed in collaboration with Xreal and built around a detachable compute puck designed to sit in a user's pocket. The WAN Show hosts were lukewarm, noting that a comparable tethered approach had not succeeded for Apple, and that the Google hardware did not, in their assessment, look as polished as Meta's competing glasses.
Pricing moved too. The hosts reported that Google's AI Ultra plan, previously priced at 250 US dollars a month, now starts at 100 US dollars a month. Free usage, however, tightened. According to the segment, free models now carry weekly hard caps plus a shorter rolling five-hour cap, and limits are calculated on compute consumed rather than on the number of prompts submitted, meaning that more complex prompts or more capable models exhaust an allowance faster.
For the advertising and marketing technology industry, the engagement gap is not a trivia point. It sits on top of a set of structural changes that directly affect how brands reach audiences.
The first is the model swap inside Search. PPC Land has tracked the erosion of organic click-through rates since SISTRIX data from March 2026 showed click-through at position one falling from 27% to 11%. When the model behind AI Mode changes, and AI Mode already serves more than a billion users monthly, the surface that marketers optimise for changes with it.
The commerce layer is shifting on the same timeline. PPC Land reported that Google launched Universal Cart at the event, a cross-merchant shopping hub spanning Search, YouTube, Gmail and the Gemini app, rolling out in the United States. According to that coverage, Google processes more than a billion shopping interactions per day, running on a Shopping Graph that the company describes as holding more than 60 billion product listings. Automated agents that monitor inventory and complete purchases - the same scalping concern the WAN Show raised - are the consumer face of that infrastructure.
The advertiser-facing changes arrived the next day. PPC Land's recap of Google Marketing Live 2026 reports that Google staged its 13th annual Marketing Live event the day after I/O, livestreaming to more than 100 countries and unveiling tools including Ask Advisor, new AI Mode ad formats, and Meridian inside Analytics 360. The throughline, that coverage notes, is Google asking advertisers to hand more of the campaign execution layer to Gemini.
There is also the question of trust, which the WAN Show hosts returned to repeatedly. If a generated visualization cannot be relied upon without independent verification, the same doubt extends to agent-completed transactions and agent-summarised data. Comments under the WAN Show video echoed this. One viewer raised the RSVP example directly, asking what happens when an agent omits a guest or records the wrong detail. For marketers building on agentic surfaces, that reliability question is not abstract: it shapes whether consumers act on AI-mediated recommendations at all.
The standards groundwork is being laid in parallel. PPC Land reported that the keynote also revealed WebMCP moving into a Chrome 149 origin trial, a proposed open web standard that lets websites expose structured tools to AI agents directly, rather than forcing agents to interpret pages through screenshots and simulated clicks. For publishers and e-commerce operators, that changes how an automated visitor reads and transacts on a site.
The WAN Show segment closed on the economics, and that thread connects the keynote to a broader industry concern. One host argued that the pricing changes - cheaper Ultra, tighter free limits, compute-based caps - suggest cost discipline taking hold after a long stretch of aggressive spending. "It does seem like the bean counters have woken up," he said.
The hosts pointed to a related signal: a reported arrangement in which Anthropic rents capacity from a data center built by Elon Musk's xAI. The figure cited in the segment was a payment of 1.25 billion US dollars a month over three years. The hosts also discussed a report, dated two days before the recording, that Anthropic was projecting its first profitable quarter, which would make it, in their framing, the first pure AI company to reach profitability. PPC Land has documented Anthropic's scale in adjacent reporting, including its 13 billion US dollar Series F round at a 183 billion US dollar valuation in September 2025. None of these specific financial claims can be confirmed here, and they are presented as the hosts described them.
The underlying point is straightforward. The volume of compute behind features like persistent background agents and on-the-fly generative interfaces is expensive, and the keynote's reception suggests a public that is watching closely without being convinced. A keynote that draws millions of viewers but almost no applause is data. For an industry being asked to rebuild its campaigns, its commerce flows and its measurement around these tools, the size of that gap is worth noting.
Who: The WAN Show hosts Linus and Luke analysed Google I/O 2026, the developer conference run by Google. The hosts assessed both the announcements made by Google leadership, including Sundar Pichai, and the public response to the keynote recording.
What: Google I/O 2026 introduced the Gemini 3.5 model family, the Gemini Omni video model, the Gemini Spark agent, an updated Xreal smart-glasses design, AI integration across Search, and revised AI Ultra pricing. Despite the volume of announcements, the keynote video drew an unusually low ratio of likes to views - cited in the WAN Show segment at roughly 0.05% of viewers - alongside a viewer-posted tally in which dislikes exceeded likes.
When: The keynote took place on May 19, 2026, with Google Marketing Live following on May 20. The WAN Show segment examining the reception was published on May 26, 2026.
Where: Google I/O 2026 was held in Mountain View, California, and streamed worldwide. The model and search changes apply across countries where AI Mode operates, while agent features such as Gemini Spark begin with United States subscribers.
Why: The muted reception matters because Google is asking the advertising and marketing industry to rebuild campaigns, commerce flows and measurement around the same AI tools the keynote showcased. AI Mode already serves more than a billion monthly users, organic click-through rates have fallen sharply, and trust in agent-generated outputs remains unsettled - making the gap between a large audience and a small show of approval a signal worth tracking.
