This is Part 2 of a 3-part series. Part 1 covered what to stop and start - by team, no fluff. This part goes deeper into the why behind each shift. Part 3 covers how SEO, Ads, and Marketing teams work together in this new structure.
Why Part 2 Exists
Part 1 gave you the what. Stop doing this. Start doing that.
But here's the thing about tactics without context - they don't stick. And they don't hold up when you're sitting across from a client whose impressions just dropped 18%, or a CMO asking why the PPC team wants to overhaul attribution mid-year, or an e-commerce director who thinks UCP "sounds like a future thing."
You need the why. Not just to convince others - but to be genuinely confident in the direction yourself.
So let's go through every major shift from Google I/O 2026 and Google Marketing Live 2026 and explain what's actually happening underneath.
Why Impressions Will Drop - Before Clicks Do
Let me start with the metric that's going to cause the most panic in the next 3–6 months.
Impressions. Specifically, impressions on informational and top-of-funnel content.
Here's what's happening: Google's new Intelligent Search Box - the biggest redesign of the search input in 25 years - now predicts and completes full search intent before a user finishes typing their query. The box expands dynamically. It suggests complete queries. It resolves vague intent into structured searches.
The result: broad, exploratory, informational queries are being completed and answered inside the box itself. The user never submits a full search. No submitted query means no impression recorded for your content.
This isn't a ranking penalty. This isn't a site issue. This is a fundamental behaviour shift baked into the product.
And here's the data that should already be on your radar: U.S. Google desktop searches per user fell nearly 20% year-over-year between 2024 and 2025, according to a Datos/SparkToro report - and that was before the Intelligent Search Box launched.[cite:118] The new search box will accelerate this trend significantly in 2026 and beyond.
The question your client will ask: "Why are our impressions falling? We haven't changed anything."
The answer: "Users are resolving their queries before they finish typing. Google answered them before the search was submitted. That's a platform behaviour shift - not a site performance issue."
What impressions will hold: specific, intent-rich, decision-ready queries. "[Product A] vs [Product B] for [specific use case]." "[Service] in [City]." "[Brand name] review 2026." These queries are precise enough that the AI can't fully resolve them with an autocomplete suggestion. They still get submitted. They still generate impressions. Those are the impressions worth fighting for.
Why Thin Content Is Now Invisible - Not Just Low-Ranking
This is the one that most SEOs haven't fully internalised yet, because the old mental model was: "thin content ranks low, we optimise to rank higher."
The new reality is fundamentally different.
Gemini 3.5 Flash is now the default model powering AI Mode in Google Search globally.[cite:114] It's not just a better ranking algorithm. It's a generative model that actively constructs AI answers by selecting which sources to pull from, synthesise, and cite.
That selection process is where thin content gets filtered out - entirely.
When Gemini 3.5 Flash generates an AI Overview or AI Mode answer, it is looking for:
- Depth - does this page go beyond the surface level?
- Perspective - does this page have a point of view, or is it just aggregating public information?
- Experience - is there evidence that a real person with real expertise wrote this?
- Specificity - does this page answer the actual question with precision, or give a generic answer?
Content that fails these filters doesn't rank #10. It doesn't get a featured snippet. It simply doesn't enter the generation pool. It is invisible to the AI layer.
The practical implication: if your content strategy has been built on publishing volume - targeting large numbers of keywords with moderately useful, largely similar content - that strategy is structurally broken now. Not weakened. Broken.
The question your team will ask: "We have 400 blog posts. Do we delete them all?"
The answer: No - but you should audit them ruthlessly. Content that has genuine depth, a clear perspective, and real specificity? Keep it and update it. Content that is generic, undifferentiated, and covers topics Gemini can answer with a two-sentence AI Overview? It's not helping you - and it may be diluting your site's overall topical authority signal.
The new content model is simple to say and hard to execute: fewer pieces, far more depth, written by or attributed to people with genuine expertise. First-hand experience, real case studies, named authors, actual opinions. E-E-A-T is no longer a checklist - it's the filter between existing in AI answers or not.[cite:115]
Why Information Agents Are a Bigger Deal Than Anyone Is Saying
Most of the coverage on Information Agents has focused on the user-facing experience: "agents that monitor the web for you." That's accurate but it misses the deeper strategic implication for publishers and marketers.
Here's the mechanism. Information Agents work proactively - they monitor websites, social platforms, news sources, and shopping outlets in the background, 24/7, and surface relevant updates to users without those users ever typing a search query.[cite:104]
Now think about what this does to the relationship between search volume and user interest.
Historically, keyword research worked on a simple assumption: if users are interested in a topic, they search for it, and that search activity shows up as measurable volume in tools like Google Search Console, Keyword Planner, or Ahrefs.
Information Agents break this assumption.
A topic can have enormous user interest - agents are actively monitoring it for thousands of users - with zero measurable search volume, because none of those users are typing queries. The agents are doing the searching for them.
This means:
- Keyword volume becomes a lagging metric - it measures what users searched for in the past, not what agents are monitoring now
- Content freshness and update frequency become primary signals - agents prioritise fresh, frequently-updated content for monitoring queues
- Structured data becomes even more critical - agents need machine-readable content to parse and surface efficiently
- Publishers without clear topical focus will get deprioritised - agents build monitoring queues around trusted, authoritative sources in specific topics
The question your content team will ask: "If people aren't searching, why are we creating content?"
The answer: "Because agents are reading it for them. The goal shifts from 'rank for the query' to 'be the source agents trust to monitor.' That's not less important - it's harder to achieve and more durable once you have it."
Information Agents are available to AI Pro and Ultra subscribers initially, rolling out more broadly over time.[cite:119] But the agents are crawling content now. The groundwork you lay today determines whether you're in their monitoring stack when they scale.
Why U.S. Search Volume Was Already Falling - And What Accelerates It Now
Before we even talk about the new features from I/O 2026, it's worth grounding this in actual data.
U.S. Google desktop searches per user fell nearly 20% year-over-year in 2024–2025.[cite:116] That's not a rounding error. That's a structural shift in how users engage with search.
And the cause, according to Rand Fishkin and the SparkToro/Datos research: users are getting what they need without running multiple follow-up searches.[cite:118] AI-powered answers and zero-click results are resolving intent faster. Users are doing fewer total searches per session.
Importantly - this happened before the Intelligent Search Box, before Information Agents, and before Gemini 3.5 Flash became the default. These features will only accelerate the trend.
What does this mean strategically?
- Total impressions across the industry will compress - this isn't your site underperforming, it's the market shrinking for certain query types
- Traffic quality matters more than traffic volume - a smaller number of highly-intent visitors converts better than a large number of exploratory ones
- Share of voice in AI citations is the new market share metric - being cited inside Gemini's answers is the equivalent of owning page 1 in 2019
Why UCP and AP2 Are Urgent for E-Commerce Right Now
Here's the framing I keep using with e-commerce clients, because it cuts through immediately:
"Amazon built a system where users shop inside Amazon. Google just built the same thing - except it works across every merchant on the internet, simultaneously."
Universal Commerce Protocol (UCP) is the technical standard that connects merchants, platforms, and AI agents into one shopping layer. Agent Payments Protocol (AP2) is the payment infrastructure that lets AI agents complete purchases on behalf of users, within defined spending limits, without requiring user intervention at checkout.
Together, they create a world where: a user tells Gemini "buy me the best noise-cancelling headphones under ₹15,000" - and the agent browses, compares, selects, and completes the purchase. Without the user visiting a single product page.
This is live in the United States today. Expanding to Canada, Australia, and the UK imminently.
The reason UCP urgency is being underplayed in most coverage is that it feels like a technical integration story - something for dev teams, not marketing teams. But the strategic consequence is a marketing story:
Merchants inside UCP = visible to agents in the purchase flow.
Merchants outside UCP = invisible to agents, regardless of how well they rank on Google Shopping.
Your SEO ranking, your Shopping campaign bid, your product page optimisation - none of it matters if the agent making the purchase decision on behalf of the user can't access your inventory through UCP.
The question your e-commerce director will ask: "How long do we have before this matters?"
The answer: "It matters today for US merchants. The merchants moving first are claiming agent-layer visibility right now. This is not a 2027 consideration."
First-mover advantage in protocol adoption is real. The shelf space inside Google's agentic shopping layer is finite and it is being occupied right now by merchants who are integrating.
Why Brand Authority Just Became a Hard Performance Signal
This is probably the most important shift for CMOs and marketing leaders to understand - and the hardest one to explain to performance-focused teams who think in CPC and ROAS.
When Google's Information Agents and AI Mode surface recommendations to users - or when Gemini constructs an AI Overview answer - they are not running a real-time keyword match. They are making trust-based decisions informed by what the model has learned about which brands, sources, and entities are credible, authoritative, and relevant.
That trust is built from signals like:
- PR coverage in credible publications - editorial mentions that confirm your brand is real and recognised
- Wikipedia presence - arguably the highest-trust entity signal the AI has
- Review platform data - Google Reviews, Trustpilot, G2, industry-specific platforms
- Community mentions - Reddit, forums, LinkedIn conversations, industry communities
- Backlink authority - who's linking to you and whether those sources are themselves trusted
- Branded search volume - when people search for you by name, it signals real-world demand and recognition
Every one of these signals is something that has traditionally lived in the "brand and awareness" budget, separate from SEO and paid performance budgets.
That separation is now a structural problem.
Because if your brand signals are weak - if Gemini's model doesn't "know" your brand from credible external references - your content is less likely to be cited in AI answers, your products are less likely to be surfaced by agents, and your ads are less likely to benefit from Attributed Branded Searches and QFCs.
Brand investment is now bottom-of-funnel infrastructure. It is not separate from performance marketing. It is performance marketing - it just operates on a longer attribution cycle.
The question your CFO will ask: "How do we measure the ROI of brand investment?"
The answer: "Through Attributed Branded Searches in Google Ads, through Qualified Future Conversions, and through branded search volume growth in Search Console. For the first time, Google has given us measurable attribution for brand impact on commercial outcomes. Use it."
Why Google Is Building a Closed Loop - And Why That's the Real Story of I/O 2026
Let me step back and name the pattern that ties every single announcement together.
Look at the full list:
- Intelligent Search Box → keeps users inside Google longer at the point of query
- Gemini 3.5 Flash in Search → resolves more queries without users clicking to external sites
- Information Agents → surfaces content for users without search queries being typed at all
- AI Mode with seamless AI Overviews flow → takes users deeper into AI conversation, away from blue links
- Universal Cart → keeps shopping inside Google across all surfaces
- AP2 → completes purchases inside Google without users visiting merchant sites
- Direct Offers → surfaces promotions inside AI search journeys
- Business Agent for Leads → handles lead qualification inside ads
- Ask Advisor → handles campaign management and reporting inside Google's tools
- Meridian in Analytics 360 → handles attribution and measurement inside Google's ecosystem
- Qualified Future Conversions → connects ad activity to future sales inside Google's data model
Every single one of these moves in one direction: less reason to leave Google.
This is not accidental. This is the deliberate architecture of a closed commerce and intelligence loop - where:
- Users don't leave to find information
- Users don't leave to shop
- Advertisers don't need external tools to manage campaigns
- Advertisers don't need external tools to measure attribution
- Google participates in every transaction, every insight, every conversion
Understanding this is what separates teams that adapt strategically from teams that react tactically to each feature announcement.
The features are not the story. The architecture is the story.
And your job - as an SEO, a paid media manager, an e-commerce lead, or a CMO - is to figure out how to remain valuable, visible, and commercially relevant inside a system that is deliberately designed to complete tasks without requiring external destinations.
What's Coming in Part 3
Part 3 is the practical one - and in some ways the hardest one to get right.
It's about how SEO, Ads, and Marketing teams work together in this new structure. Not the theory. The actual operating model: what signals each team generates, how those signals feed the next team's performance, and what a unified intelligence loop looks like in practice.
It also covers how to have the conversation with leadership - the framing that holds across all three teams and doesn't require a 40-slide deck to explain.
More signal, less noise. Always. 🙏
Sources referenced in this post: Google I/O 2026 official announcements · GML 2026 Google Blog · Gemini 3.5 Flash announcement · Gemini 3.5 Flash on TechCrunch · Google Search I/O 2026 updates · Intelligent Search Box - Search Engine Land · Information Agents - Search Engine Land · Universal Cart + AP2 - Search Engine Land · Google searches per U.S. user fell 20% - Search Engine Land · Google searches drop 20% - MarTech · Meridian in Analytics 360 · GML 2026 full recap - Search Engine Land
.jpg)
Good readers always drop comments!!