Our Most Important Learnings From Google I/O and Marketing Live 2026
22 May 2026
The conventional reading of this week's two Google keynotes is that Google shipped a lot of AI features. That is true but misses the point. The more consequential thing that happened is structural: Google is redesigning the fundamental input-output relationship between marketers and the platform.
Keywords are becoming one input among many. The new Search box, AI Mode, and background search agents are changing how queries form before they ever reach your campaigns. AI Max and the new ad formats are changing how ads match those queries. And what your website says, how your conversion data flows, and how clearly your brand communicates its purpose are becoming the actual targeting levers.
This article covers both events in the sequence that makes the most sense: first what changed in Search, then the ad product built for that new Search, then the website architecture that underlies both, then YouTube's measurement story, then the infrastructure shift that makes all of it technically possible.
Google Search Is AI Search, Structurally and Not Just in Name
AI Mode has been live for just over a year. In that time it passed one billion monthly users, and searches made through it have been doubling every quarter since launch. AI Overviews now reaches 2.5 billion monthly users. Total Search queries hit an all-time high last quarter.
The pattern is consistent: when Google makes Search more capable, people use it more. And the character of what they ask is changing. AI Mode searches are on average three times as long as traditional searches. Brainstorming queries, searches using terms like "ideas for" and "which should I", are growing 30% faster than AI Mode searches overall. Users are arriving with complex, multi-part questions that previously required ten open tabs.
Google announced the biggest upgrade to the Search box in 25 years. It now expands as you type, adds AI-powered nuance to your query, and accepts text, images, files, and video as simultaneous inputs. The system helps users ask the question they actually have, not the compressed keyword they previously typed.
Two further capabilities are launching this summer. Search agents will run 24/7 in the background, monitoring topics without an active query, and surfacing relevant updates when conditions are met. Generative UI, built on Gemini 3.5 Flash and Antigravity's agentic coding platform, will construct custom interactive layouts in real time, tailored to the specific question being asked rather than serving a static format.
For marketers, the implication is direct. A keyword set built around two-to-three word phrases is structurally misaligned with how users now interact with Search. The signal of intent is no longer the keyword. It is the full context of the query, the AI-generated response, and the user's conversation history.
AI Max Is the Ad Product Built for This Environment
Google introduced AI Max for Search at last year's Marketing Live. It is now out of beta and available globally. The reported performance: 27% more conversions compared to manual search campaigns, a 14% average uplift in conversions at similar CPA or ROAS when enabled, and for campaigns previously reliant on exact and phrase match, a 27% uplift specifically (Google Marketing Live, May 2026).
AI Max is not a new campaign type. It is an optimisation layer applied to existing Search campaigns, extending what Search can do by giving Google's AI control over three things: how search terms are matched to your ads, what creative is served, and which landing page on your site the ad points to.
The three pillars work independently but compound together. Search term matching uses broad match and keywordless technology simultaneously, learning from your keywords, creative assets, and URLs to expand into relevant queries your keyword set does not explicitly cover. Text customisation uses extractive and generative AI to create additional headlines and descriptions grounded in your landing page content. Final URL Expansion allows Google to replace the URL in the ad with a more relevant page on your domain, based on the user's query.
Each pillar can be toggled on or off. But the performance case for enabling all three is strong. Across 23 tests involving 16 mature advertisers, Search Engine Land found 40% higher test success rates when all three features were active versus search term matching alone.
The new ad formats announced at Marketing Live are exclusive to AI Max and Performance Max. Direct offers deliver personalised deals dynamically generated by Gemini from your product and promotional rules, now available inside AI Mode conversations. Business Agent for Leads shows an interactive unit where users can ask questions answered directly from your website assets, with pre-filled lead forms at the end. AI-powered shopping ads surface in conversational search results with product features written in natural language. Standard Search campaigns are not eligible for any of these formats.
The control mechanism for AI Max is AI Brief, announced at the event. It replaces the text guidelines product that preceded it, with existing guidelines carrying over automatically. AI Brief steers AI Max through natural language across three categories: messaging guidelines covering what ads should and should not say, matching guidelines covering which searches to pursue or avoid, and audience guidelines covering how to tailor messaging for different segments. It provides previews before committing. It is currently in English for AI Max Search, with Performance Max and AI Max Shopping support following.
A hard deadline applies. In September 2026, Dynamic Search Ads are being retired. DSA campaigns will migrate automatically to AI Max with all three features enabled by default. DSA page-condition rules cannot be recreated in AI Max after migration. The time to prepare the account architecture is now, not late August.
Your Website Is Now Part of the Campaign Architecture
The buried headline of AI Max is straightforward: your website is not just a post-click destination. It is a targeting input.
Text customisation draws from your page titles, meta descriptions, and body copy. Final URL Expansion draws from your page structure and content to select which URL serves which query. Keywordless matching interprets what your business does from what your pages say. Google's AI reads your site and uses that reading to decide what to show, and to whom.
The practical consequence: if your pages are vague, outdated, or legally unsafe, AI Max inherits those problems and amplifies them at scale. Thin category pages produce thin, generic ads. Pages that do not clearly communicate who they serve, what problem they solve, and what action the user should take are weak inputs. The system's output reflects that directly.
The September migration creates urgency for anyone currently running DSA. At minimum, before AI Max is active: a complete URL inventory with exclusions for pages not suitable for conversion traffic, page feeds with custom labels to control which pages are eligible for which intent clusters, text guidelines covering brand voice and compliance requirements, and conversion tracking configured at depth rather than surface level. Form submissions alone are not sufficient signal quality for AI Max to optimise toward good business outcomes rather than cheap leads.
This is also where the I/O announcements connect to the advertising story. Search agents running continuously in the background will surface brands already in Google's consideration set for a category. Website architecture, content clarity, and internal linking structure directly influence whether AI Max and, in time, search agents serve your ads in the right contexts. SEO and paid search are no longer separate functions.
YouTube's Performance Case Is Getting Harder to Ignore
Google's ROAS numbers for YouTube have been moving consistently in one direction. A 2026 media mix model study with TransUnion found YouTube drives 40% higher ROAS than all other media combined, up from 21% the previous year. Long-term ROAS is more than double that of linear TV, paid social, and streaming. Against paid social specifically, YouTube's long-term effectiveness is 86% higher (Google Marketing Live, May 2026).
The audience reach is specific, not generic. 45% of YouTube Shorts viewers are not on TikTok. 65% are not on Reels. YouTube reaches 90% of US adults. It has been the number one streaming platform by watch time for three consecutive years, and two billion Shorts are watched on connected TVs each month.
Demand Gen gained several concrete updates. Placements are expanding to Google Maps, where users are already in an active purchase mindset. Product feeds, which already drive 33% more conversions when used, are extending to tablets and to pause ads on mobile. Adding creator assets to Demand Gen campaigns increases conversion lift by 20% on average (Google Marketing Live, May 2026). The ability to boost affiliate creator videos directly from Demand Gen is now available.
Two measurement updates address the persistent problem of proving YouTube's contribution. Campaign type attribution now shows every conversion Demand Gen contributes to alongside other channels, without the deduplication that previously distorted the picture. Qualified Future Conversions (QFC) connects upstream signals, including branded searches, video views, and site visits, to downstream conversions up to six months later. UK retailer Crew Clothing ran a YouTube prospecting campaign that appeared flat over 30 days. Through QFC, a 70% uplift in long-term conversions was visible. They scaled the campaign instead of cutting it.
Meridian, Google's open-source marketing mix model, is now integrated into a redesigned Google Analytics 360 that measures across Google, TikTok, Pinterest, Snap, and other platforms from a single interface, with cross-channel scenario planning built in.
The Infrastructure Shift That Makes All of This Technically Possible
The I/O keynote explains why these advertising capabilities exist now and did not two years ago. Google is processing 3.2 quadrillion tokens per month, up from 480 trillion a year ago: a 7x increase in twelve months. Capital expenditure is tracking at $180 to $190 billion this year, six times the $31 billion deployed in 2022. The eighth generation of TPUs includes specialised architectures for training and inference, with the inference chip capable of generating approximately 1,500 tokens per second.
Gemini 3.5 Flash is the model running AI Mode, AI Max's matching logic, the new ad formats, and Asset Studio's creative tools. It operates at four times the output speed of comparable frontier models at less than half their price. For AI Max to deliver custom creative and custom landing pages in milliseconds across billions of daily queries, that speed is the prerequisite, not a feature.
Gemini Omni, announced at I/O and coming to Asset Studio this summer, combines Gemini's reasoning with the best generative media models to produce video, image, and interactive content from any input combination. Creative drives nearly half of incremental ad sales according to Google's own research (Google Marketing Live, May 2026). The tools for producing that creative are compressing production cycles from days to minutes.
For shopping, the Universal Commerce Protocol (UCP) is now co-steered by Amazon, Meta, Microsoft, Salesforce, and Stripe alongside Google. The Universal Cart aggregates products across merchants and Google surfaces, tracks price changes, applies payment card perks automatically, and supports direct checkout from ads for retailers with UCP configured. The infrastructure for agent-mediated commerce is live, not theoretical.
Antigravity 2.0, Google's agentic development platform, is worth noting for the signal it sends. Using Gemini 3.5 Flash and Antigravity agents working in parallel, Google built a functioning operating system from scratch in 12 hours for under $1,000 in API costs. That figure illustrates what frontier intelligence at Flash's price point makes feasible. The same agentic coding capability is now inside Search, building custom interactive UI responses for specific queries in real time.
The Shift in One Sentence
Reading both keynotes together, the picture is consistent: the quality of your inputs, your website, your creative, your conversion signals, your brand clarity, determines your outcomes more than your manual campaign management does. The shift is not from human to machine. It is from human operating levers to human configuring inputs, with the system doing the real-time optimisation.
For accounts relying on DSA, September 2026 is a hard migration date and the window for preparation is narrowing. For Search campaigns broadly, AI Max readiness is a strategic decision about account architecture, website quality, and conversion tracking depth, not a feature toggle. Getting those foundations right before the migration is what separates accounts that benefit from the transition from those that get dragged through it.
Kaliber works with clients across Southeast Asia on exactly this: auditing current campaign architecture against AI Max requirements, building page feed structures, writing text guidelines, and aligning conversion tracking to business outcomes rather than form submissions. If your account is heading into September without a clear AI Max plan, the starting point is here.
Frequently Asked Questions
Is AI Max available globally right now?
Yes. Google confirmed at Marketing Live 2026 that AI Max for Search is out of beta and available globally. AI Max for Shopping was announced on 30 April 2026 and is also available as a one-click upgrade for existing Shopping campaigns. AI Max for Travel, consolidating hotel ads, flight ads, and general travel search into one campaign type, was also announced and is now eligible to appear inside AI Overviews and AI Mode.
Do I need to restructure all my Search campaigns before September 2026?
Not necessarily restructure, but you do need to audit and prepare before the automatic migration. DSA page-condition rules are not supported in AI Max and cannot be recreated after conversion. At minimum: URL exclusions blocking unsuitable pages, page feeds with custom labels, text guidelines covering brand voice and compliance, and conversion tracking configured at depth beyond top-of-funnel form submissions. Accounts that wait for Google to apply the automatic migration will lose the ability to shape how their legacy DSA logic translates.
What is AI Brief and is it different from text guidelines?
AI Brief is the successor to text guidelines, announced at Marketing Live 2026. It allows advertisers to steer AI Max through natural language across three categories: messaging guidelines, matching guidelines, and audience guidelines. Existing text guidelines carry over automatically. AI Brief is currently in English for AI Max Search and will expand to Performance Max and AI Max Shopping. The key addition over text guidelines is the matching and audience dimensions: you can now tell the system which search contexts to pursue or avoid, and how to adapt messaging for different audience segments.
What is happening with keywords if keyword lists are no longer the primary targeting lever?
Keywords remain an input into AI Max but are one input among many. Under AI Max, effective targeting is encoded in the combination of keyword themes, ad copy, landing page content, site structure, page feed labels, URL controls, brand lists, text guidelines, conversion signal quality, and bid strategy. Campaign structure should be organised around intent clusters and business decision boundaries rather than individual keyword variations. The search terms report in AI Max distinguishes between keyword-led expansion and keywordless landing page-based matching, which is the signal that tells you how much of your coverage is keyword-derived versus page-derived.
What does the YouTube ROAS data actually measure, and is it credible?
The 40% higher ROAS figure comes from a 2026 media mix modelling study conducted with TransUnion, presented at Google Marketing Live 2026. MMM methodology attributes revenue to channels across the full customer journey rather than relying on last-click or single-touch attribution. The 86% long-term effectiveness advantage over paid social uses the same methodology. These figures are sourced from Google and TransUnion jointly, and the long-term ROAS metric specifically captures conversion value that occurs beyond the initial 30-day attribution window. Crew Clothing's 70% long-term conversion uplift, also presented at the event, is a named brand with a specific metric, not an anonymised case study.
