Google I/O 2026 marked a turning point, not an incremental upgrade. AI is no longer a feature layer on top of search; it has become the primary way people discover, evaluate, and act on information. Here is what changed, why it matters, and how to respond.
Google now processes over 3.2 quadrillion tokens per month, a 7x increase year over year. AI Mode in Google Search crossed one billion monthly active users. Zero-click searches account for 72% of AI Overview sessions. These numbers are not projections; they describe the web as it exists today.
The keynote introduced Gemini 3.5 Flash, a model that combines reasoning with autonomous action and benchmarks at the frontier while costing half of comparable alternatives. Alongside it came Gemini Spark, a 24/7 background agent that monitors topics and surfaces synthesized answers before a user types a single query. The biggest redesign of Google’s search interface in 25 years replaced keyword matching with a conversational, multimodal experience that accepts text, images, video, files, and open browser tabs simultaneously.
For brands, the practical consequence is this: 88% of AI Mode citations come from sources outside the organic top-10 search results. Traditional ranking signals alone will not determine which brands appear in AI-generated answers. Discoverability now depends on how well your digital presence is structured, consistent, and machine-readable.
Google’s new frontier model not only answers questions but executes multi-step tasks on behalf of users. Gemini Omni handles video generation from any input, whether text, image, audio, or video, with physics-aware rendering and built-in SynthID watermarking. The practical upshot: AI that acts is now cheaper and more capable than AI that only answers.
The Google Search interface was rebuilt from the ground up. The new Intelligent Search Box accepts text, images, uploaded files, streaming video, and open browser tabs simultaneously, moving queries from keywords into multi-turn conversations. AI Mode, already at one billion users, generates synthesized, cited answers rather than a list of links. Gemini 3.5 Flash powers these responses using retrieval-augmented generation, pulling specific facts directly from the search index rather than rendering full web pages. The ten blue links have not disappeared, but they are no longer the primary interface.
Universal Cart is the most commercially significant announcement for businesses that sell anything online. One intelligent cart operates across Google Search, Gemini, YouTube, Gmail, and Google Maps. It actively monitors price drops, tracks stock availability, and uses the Universal Commerce Protocol (UCP) for cross-merchant checkout. This means people can convert from anywhere, without visiting a website. Booking and Expedia are already integrated into AI-assisted trip planning through this protocol, surfacing direct offers inside itinerary flows. Google also introduced the Agent Payments Protocol (AP2) to handle secure cross-merchant transactions at scale.
Gemini Spark operates as a continuous background agent, running even when a user’s device is off, monitoring user-defined topics and pushing synthesized updates proactively. Information Agents embedded in Google Search do the same at the platform level, continuously scanning websites and cross-referencing information without waiting for a user query. This is not a periodic crawl. It is persistent, real-time monitoring of your entire digital footprint.
The move to AI does not retire the fundamentals. Strong SEO still matters, and the signals Google rewards in AI are the ones it has pushed for years: Experience, Expertise, and genuinely Helpful Content. Brands that built authority and answered real questions are the easiest for AI to summarize, cite, and recommend. What changes is the surface. Advertising shifts to a “goal-in, AI-executes” model where “the best ads are answers,” matched on the full context of a search rather than isolated keywords. AI Mode queries run longer and more specific, so the system reads intent, not terms.
Ads in AI Mode sit inside the conversation rather than above or below it, surfacing as the next logical step in a user’s follow-up questions. Google is testing two formats: Conversational Discovery ads, where Gemini builds creative in real time from your assets to fit the query, and Highlighted Answers, a clearly labeled sponsored option inside an AI recommendation list. In both, Gemini writes in your brand voice, and that voice stays brand-controlled, not rewritten by the model.
Direct Offers in AI Mode drop brand offers into the response at the moment of purchase intent: coupons and promotions for retail, rate offers for hoteliers. Live since January 2026 with Chewy, Gap, and L’Oreal, the format lets Gemini bundle promotions per query, and a “buy with one-time code” button applies the discount and checks out through Google Pay.
AI-Powered Shopping Ads use Gemini to explain products in a natural, conversational tone. Search “espresso machine” and Gemini surfaces your most relevant products with a custom explainer for each, not a recycled feature list.
Business Agent for Leads puts a Gemini agent inside the ad. Instead of a static form, a prospect clicks “Chat” and gets instant answers grounded in your website, qualifying intent before handoff to sales. Google is testing it first in education, automotive, and real estate.
AI Brief replaces keyword lists with plain-language control across three areas: messaging (what to say and avoid), matching (which queries to win or exclude), and audience (tailored messages per segment). Set the guardrails, and the system matches fluid intent within them, with asset and search previews before launch.
AI Max and Performance Max are now built for every advertiser and form the foundation the new AI Mode formats run on. Google reported adopters saw roughly 15% more conversions, the payoff for giving the system clean inputs and clear goals instead of managing execution by hand.
On Commerce:
Universal Cart becomes the agentic hub for shopping on Google. Users add to cart while browsing or chatting across Search, Gemini, and YouTube, and the cart tracks price drops and back-in-stock alerts until they are ready to buy. Native Checkout through the Universal Commerce Protocol (UCP) closes the loop on Google itself, moving users from discovery to discounts to conversion through Google Pay, backed by a Shopping Graph of more than 60 billion listings. The same protocol powers YouTube Ads, putting the full discovery-to-conversion journey inside YouTube. UCP expands to Canada, Australia, and the UK in the coming months, and beyond retail into travel, with Booking and Expedia surfacing offers in AI-assisted trip planning.
The throughline: success now depends far less on keyword management and far more on feed completeness, clean first-party data, structured content, and clear strategic inputs. The same SEO disciplines of expertise and helpful content feed the machine that now powers paid placement and agentic checkout.
Google released an entire suite of agentic commerce protocols to make agentic commerce easier, including WebMCP, MCP, the Agentic Discovery framework, UCP, and more.
The proposed WebMCP standard (Model Context Protocol for Web) formalizes how websites expose their functionality to browser-based AI agents. Forms, JavaScript functions, and API endpoints can be declared in a machine-readable format so that agents understand not just what a page contains, but what actions it supports. Google’s developer sessions formalized a nine-dimension Agentic Discoverability framework as the practical benchmark for whether a site is ready for the agentic web. The dimensions cover AI crawler access, llms.txt presence, structured data quality, semantic HTML, sitemap completeness, metadata, content clarity, API discovery, and contact signals.
The 72% zero-click rate in AI Overview sessions is the current state of AI-powered search. A brand can be cited as an authoritative source in thousands of AI-generated answers in a week and see flat or declining traffic in Google Search Console. The metrics that indicate brand health in the AI era, such as citation frequency, recommendation rate, and share of voice in AI responses, do not appear in any standard analytics dashboard.
This creates a real measurement problem. Marketing teams are being evaluated on metrics that no longer capture the full picture of digital performance. A brand can be winning in AI search and look like it is losing in traditional reporting, or vice versa. Getting this right matters for budget decisions, channel strategy, and demonstrating performance to leadership.
The right response is to add AI visibility tracking alongside traditional reporting, not to replace one with the other. Run your core brand and category queries through Google AI Mode. Document which sources are cited, how often your brand appears, and where competitors are being recommended instead of you. That baseline is more informative than your current ranking report for understanding where you actually stand.
How Milestone helps: Milestone GEO Intelligence Platform tracks brand presence, citation frequency, and share of voice across AI-powered search experiences, providing the visibility measurement layer that standard analytics tools do not cover. It incorporates the nine-dimension Agentic Discoverability scoring from Google’s developer guidance into a dashboard that gives teams a specific, actionable improvement roadmap.
Universal Cart changes the commerce funnel in a way that has no precedent in the past decade of digital marketing. A consumer can discover a product on YouTube, compare it in a Gemini conversation, check local availability in Maps, and complete the purchase, all within Google’s ecosystem. The brand’s website may never enter the picture.
For hospitality brands, the stakes are immediate. Booking and Expedia are already in the UCP ecosystem, surfacing direct offers inside AI-assisted trip planning. A traveler researching a destination in Google AI Mode can see an OTA offer, complete a booking, and receive a confirmation, without the hotel’s website being visited once. The rate parity and direct-booking strategies that brands have managed carefully for years are being renegotiated at the protocol level.
The practical priority is product and inventory feed quality. In a UCP-powered commerce environment, your feed is your storefront for the AI layer. Rates, availability, room types, amenities, and product attributes must be real-time and complete, not because they affect rankings but because the AI making a purchase decision on a consumer’s behalf will only surface options it can fully verify and transact on. Incomplete or stale feeds do not underperform in this environment; they are invisible.
How Milestone helps: Milestone Local keeps your location data, rates, and business attributes accurate and synchronized across authoritative channels, including maps and directory ecosystems. Your local data should be enriched with your customer data so it is real, relevant, and up to date, because accuracy, relevancy, and freshness are essential for winning in AI-powered local search. The goal for multi-location brands should be to combine the foundation of UCP-based discovery and local recommendations for AI search.
The data from I/O 2026 is direct: 88% of AI Mode citations come from sources outside the organic top-10 search results. Traditional SEO ranking and AI citability are largely independent variables. What determines citation eligibility is structural: how content is marked up and what kind of schema markup is used, how specifically it answers a question, how authoritatively it covers a topic, and how consistently information appears across the web.
Google specifically called out the following four types of schemas as being essential for discovery:
AI agents read websites through three layers: the visual screenshot, the HTML/DOM structure, and the Accessibility Tree. A page that works visually but is built on non-semantic HTML with generic class names and JavaScript-gated interactions is readable to humans and opaque to machines. The same content structured with proper landmark tags, schema markup, and descriptive attribute patterns is immediately actionable to an AI agent.
Two technical steps that pay immediate dividends: deploy JSON-LD schema for Organization, WebSite, FAQPage, and any product or service entity relevant to your business, and create an llms.txt file, a plain-text document at your site root that gives AI agents an annotated map of your content and key URLs. The file takes less than an hour to create and signals clearly that your site is built for the agentic web.
How Milestone helps: Milestone Schema Manager automates the deployment of deeply nested, dynamically updated schema across enterprise site portfolios, the structured data layer that determines whether AI can understand, cite, and act on your content.
Like so many LLMs, Gemini 3.5 Flash also uses retrieval-augmented generation to extract specific answers directly from source content. Simultaneously, generative UI capabilities allow Google to assemble custom response formats such as comparison tables, dashboards, and curated lists on the fly from cited sources. Content that summarizes what other pages say is the raw material for the AI layer, not a citation candidate within it.
The content that earns inclusion in AI-generated answers has recognizable characteristics: it answers a specific question in the opening paragraph without preamble, it reflects genuine expertise or primary research that a generalist model cannot reconstruct from other sources, and it covers a topic with enough depth that it becomes the definitive reference rather than one of many similar pages. This is a different writing discipline from keyword-optimized content, and most brand content libraries were not built with it in mind.
The same principle extends to paid formats. Ads in AI Mode and AI-Powered Shopping Ads reward brands with rich, complete, well-structured product and property data. A resort that has populated every relevant amenity, service specialty, and operational attribute in its Merchant Center feed is more eligible for these placements, and will perform better in them, than one relying on paragraph-form descriptions buried on an About page. The AI Brief model also means that strategic clarity about brand positioning now directly influences how Google systems interpret and match queries. Vague or incomplete briefs produce vague matches.
How Milestone helps: Milestone AI Content Studio generates and optimizes content for generative engine discovery, structuring depth, authority signals, and topical specificity to meet the criteria large language models use when selecting sources to cite in AI-generated answers.
Gemini Spark and Google’s Information Agents do not visit your website the way a user does. They scan continuously, cross-reference your information against other sources, make trust assessments based on consistency, and build a persistent model of what your brand offers. Inconsistencies across your website, local listings, and third-party directories are not just a UX problem. They are a trust signal failure that causes AI agents to deprioritize your brand.
As agentic commerce matures, the difference between a website that agents can transact with and one they can only read will become a meaningful competitive gap. The Gemini API and Google AI Studio now offer businesses a practical starting point for custom agent development to begin closing this gap.
How Milestone helps: Milestone CMS serves as the single source of record that keeps business information consistent across all web properties and connected channels, the consistency that helps AI agents evaluate which sources to trust and cite.
How Milestone helps: Milestone Agent Platform enables organizations to design, deploy, and govern custom AI agents and supports agent-to-agent communication as the WebMCP and UCP standards mature.
The temptation after an event like I/O 2026 is to treat the implications as a long-term planning item. The numbers argue against that instinct. One billion AI Mode users, 72% zero-click, Universal Cart already live with major OTA partners: these are not signals that the agentic web is approaching. They confirm it has arrived.
The brands that will look back on this moment as an advantage captured are the ones that moved on the structural foundations first: consistent data across channels, machine-readable content architecture, AI visibility measurement, and feed quality that supports commerce outside their own website. Most of these are known problems. The urgency is new.
Start with an honest audit across five questions:
Each gap compounds the others. Closing them in sequence builds the foundation that agentic discoverability, AI commerce participation, and sustained brand visibility are built on.
Cote Hospitality has proudly managed iconic resorts for over a century, bringing memorable experiences to guests…
Milestone announces a strategic partnership with Candescent to drive growth for financial institutions. As these…
AI is no longer enhancing search -- it is mediating discovery, evaluation, and transaction decisions, making website visits optional…
Milestone Inc. is thrilled to announce its recognition at the 2025 Hospitality Sales & Marketing…
Milestone has once again demonstrated its leadership in digital innovation, securing six prestigious WebAwards, including the coveted Outstanding…
We are thrilled to announce that Milestone has been honored at the 20th Annual W3…