TLDR: On 19 May 2026, Google announced the biggest changes to Search in 25 years: AI Mode is now the global default, the search box has been rebuilt for conversational AI prompts, persistent background information agents are rolling out this summer, and Universal Cart agentic checkout is expanding into the UK. For marketers, this has two immediate consequences. First, the measurement model you’ve been using is now structurally broken. Information agents and agentic checkout create conversions with no session, no UTM, and no trackable click path. Second, ranking number one in Google no longer means you appear in the answer. The rules for SEO visibility have changed.
Search as you know it is over. Here’s what replaced it
On 19 May 2026, Sundar Pichai confirmed at Google I/O what the data has been signalling for 18 months: Google Search is no longer a search engine in the traditional sense. It’s an intelligent assistant, a personalised research tool, and for brands increasingly the surface where purchasing decisions get made without a visit to your website.
Five announcements. All five of them affect how you measure, how you optimise, and how you plan.
What Google actually announced: the five changes that matter
The search box has been rebuilt, for the first time in 25 years
Google’s new intelligent search box expands as you type, accepts long conversational prompts, and surfaces shortcuts to AI Mode, image and document uploads, and Gemini Live. It’sdesigned to pull users deeper into AI Mode from the first keystroke, making conversational AI the default path, not an optional extra.
AI Mode is now the global default
Previously a US-only opt-in experiment, AI Mode, powered by Google’s new Gemini 3.5 Flash model, is now live for every Google user worldwide. Every UK brand that has been treating AI search as a future problem just had their timeline pulled forward by 12 to 18 months.
Information agents that run 24/7, without a new search
Users can set up persistent background agents that monitor topics indefinitely: price drops, product launches, property listings, sports results. The agent is set once and runs continuously. When it triggers a purchase, there is no search session, no click path, and no clean attribution trail back to the original moment of intent. Your measurement model has to be able to account for that. Most can’t.
Universal Cart and agentic checkout coming to the UK
Google’s Universal Cart, built on the Universal Commerce Protocol co-developed with Shopify, is expanding from the US into Canada, Australia, and the UK. AI agents can complete purchases on a shopper’s behalf inside Google itself. The user never visits the brand’s website. For marketers, this means sales will increasingly arrive without a corresponding session in your analytics. Your assisted-conversion logic was built for a world where the buyer visits your site. That world is being deprecated.
AI Overviews now keep users inside Google
Follow-up questions are now supported inside AI Mode. Users can move from initial query to research to purchase intent without generating a single referral click to a publisher or brand site. The full user journey now happens inside Google.
Gemini Omni was launched: Google’s Nano Banana moment for video
Announced at I/O 2026, Gemini Omni combines Gemini’s intelligence with Veo, Nano Banana, and Genie into a single model that generates and edits video from almost any input, text, images, audio, or existing footage. The key shift is how you edit: entirely through natural language, with each instruction building on the last. Characters stay consistent, physics holds up, and the scene retains continuity across multiple edits. Users can also generate videos using their own digital avatar, and every output is watermarked with Google’s SynthID for AI verification.
Who can use it now
Gemini Omni Flash is live today for Google AI Plus, Pro, and Ultra subscribers via the Gemini app and Google Flow and rolling out to YouTube Shorts and YouTube Create at no cost this week.
The numbers behind the shift
Before I/O 2026, here is where organic search already stood:
- 25% of all Google searches now trigger an AI Overview, rising to 48% for informational categories.
- 0.61% organic CTR on queries with an AI Overview in 2025, down 65% from 1.76% in 2024, before recovering to 2.4% by February 2026.
- 17–54% citation overlap between organic top 10 results and AI Overview sources in early 2026. Ranking number one no longer guarantees you appear in the answer.
- 190x less referral traffic sent by ChatGPT Search versus traditional Google, despite handling an estimated 12% of Google’s query volume.
Now add the I/O 2026 announcements on top of that. The intelligent search box pulls users deeper into AI Mode from the first keystroke. Information agents quietly satisfy ongoing intent, price monitoring, product tracking, sports scores, without those users ever running a fresh search. And AI Mode follow-ups keep the entire research journey inside Google, rather than releasing users to brand and publisher sites.
The uncomfortable headline: ranking number one in 2026 doesn’t mean what it meant in 2023. Visibility now means being cited inside the AI answer. The rules for that are not the same as the rules for ranking.
Why your measurement model is now broken and what to do about it
Attribution has been quietly degrading for two years. I/O 2026 just accelerated the timeline. Three of yesterday’s announcements directly degrade the data marketers rely on today:
- Information agents don’t generate sessions. When an agent monitors a topic for weeks then fires a notification that converts, the first touch was months ago in a conversation. There’sno UTM. No session. No trackable path back to which channel created that intent.
- Universal Cart removes the website visit entirely. Sales arrive without a corresponding analytics session. GA4 has no record of the journey that led to them.
- AI Mode follow-ups keep the entire research journey inside Google. Users form intent, evaluate options, and move toward purchase without generating a single referral click to your site.
The result is that your highest-converting traffic source in 2026 is also your least visible one. And if you can’t measure where revenue is actually coming from, you cannot make confident decisions about where to invest the next pound.
This is where Marketing Mix Modelling becomes the practical answer, not a nice-to-have
Last-click attribution was already an incomplete model before yesterday. Information agents and agentic checkout just finished it. MMM takes a statistical view of all your marketing activity,paid, organic, offline, seasonal, and credits the channels that drove original intent even when the final touchpoint is invisible inside a Google agent. It accounts for the surfaces you can’t track directly: AI Mode, Gemini, ChatGPT, Perplexity, Copilot.
We’ve seen what this kind of analysis reveals. Sigma Sports identified an £80,000 revenue opportunity in a single MMM analysis with Modo25. The gap between what last-click reports and what MMM surfaces isn’t marginal; it’s the difference between knowing where your marketing is working and guessing. For most brands running mixed channels in 2026, that gap is getting wider every quarter.
What this means for SEO: start optimising to be cited
The shift in how AI Overviews source their answers changes the job of SEO fundamentally. Citation overlap with the organic top 10 has dropped to between 17% and 54% in early 2026. That means a significant proportion of what appears in AI answers is not drawn from the pages that rank highest. You can rank well and be invisible in the answer. You can rank lower and be cited consistently. The two are no longer the same thing.
Here’s what the brands getting this right are doing differently:
- Leading every section with a clear, direct answer in the first 100 words. AI Overviews extract discrete, citable claims. Pages that bury the answer in a long narrative rarely get pulled into the answer.
- Using subheadings that match how a human would phrase the question. Conversational structure earns citation in a way that keyword-optimised headers often don’t.
- Adding proprietary data, original research, or a genuinely contrarian view, anything an LLM can’t synthesise from a dozen other sources. Generic content becomes structurally invisible in AI Search.
- Treating schema, structured data, and page speed as citation signals as well as ranking signals. Technical foundations now affect AI visibility, not just organic rank.
- Building brand authority outside Google. Digital PR, podcasting, YouTube, Reddit presence, community, these create the third-party signals that LLMs use to assess credibility. Brands with strong off-platform authority are consistently more cited than those relying solely on on-page optimisation.
And measure differently. If organic sessions is still your headline SEO KPI, you are measuring the wrong thing. The metric that matters is whether your brand is cited inside AI answers for the queries that matter to your business. Add citation tracking, branded search lift, and direct traffic correlation alongside traditional rank and traffic reporting.
What to do in the next 90 days
Add citation tracking to your SEO reporting
Start tracking which queries in your category trigger AI Overviews, which brands get cited, and which content formats are winning citation. This is the new share-of-voice metric. Your SEO agency should be building this into standard reporting now, if they aren’t, ask for it.
Audit your content for AI answerability [h3]
Work through your highest-traffic pages and apply a simple test: does this page lead with a clear, direct answer to the query it’s targeting? If the answer is buried in paragraph four, rewrite. AI Overviews extract the most accessible, citable claim, not the most thorough treatment of a subject.
Move measurement from last-click to MMM [h3]
This is the most important structural change available to any marketing team in 2026. Last-click can’t see the impact of brand campaigns, offline activity, upper-funnel content, or, now, AI-mediated conversions. MMM can. If you haven’t run an MMM analysis in the last 12 months, you are making budget decisions on incomplete data. Start the conversation now, before it shows up as a quarterly miss.
Build brand equity as a structural defence [h3]
Publisher sites with strong brand recognition lost 22% of search referrals in two years. Smaller sites with weaker brand recognition lost 60%. The difference is brand. When users know your name, they search for you directly, subscribe to your email, follow your social channels, and find you through multiple touchpoints. Brand-led demand is now a structural defence against Google’s narrowing referral funnel. This is the moment to invest in digital PR, video, and community in a way that may have felt like ‘top of funnel’ indulgence before.
The Modo25 view
We have spent the last six months rebuilding our SEO and measurement frameworks around exactly this. If you want to understand what your category looks like inside AI Mode today, what gets cited, what doesn’t, where your measurement has blind spots, and what your competitors look like in an AI Overview, get in touch with Modo25 today.
