Skip to content

Chrome’s AI Mode Upgrade: How Side-by-Side Search Reshapes Brand Visibility, Commerce, and the Click Economy

Google rolled out a significant update to AI Mode in Chrome. The announcement came through the official Google blog post on AI Mode in Chrome, and the changes are live now for US users on desktop and mobile.

The headline feature is side-by-side browsing: when you click a link inside AI Mode on Chrome desktop, the webpage opens in a split-screen view alongside the AI conversation panel. Research context stays on one side while the source page opens on the other. The AI also uses the open page as context for follow-up questions.

The second feature is cross-tab search: a new “plus” menu on the New Tab page and inside AI Mode lets you select recent tabs, images, and PDFs, then bring that context into a single AI Mode query. You can combine a product page, a PDF spec sheet, and a competitor tab and ask the AI to compare them.

These might sound like incremental UI improvements. They are not. They represent a structural shift in how information flows from publisher to user, with cascading consequences for brand visibility, programmatic advertising, and the entire commerce funnel.

Contents

The Browser Is No Longer a Window. It Is a Workspace.

Chrome’s AI integration has been building in stages throughout early 2026:

January 2026: Google introduced the Gemini side panel in Chrome, powered by Gemini 3, with Auto Browse - an agentic feature that lets Gemini navigate sites, add items to carts, and complete reservations on the user’s behalf. Google detailed this in its Gemini 3 Auto Browse announcement and in the broader Gemini in Chrome overview.

April 14, 2026: Chrome added a Skills library - users can save reusable AI prompts (like “compare protein macros for this recipe” or “generate side-by-side spec comparisons across these tabs”) and replay them with a single click on any website. Industry coverage treated this as a quiet but significant step because it turns casual AI queries into persistent, reusable workflows.[ALM guide]

April 16, 2026: The side-by-side browsing and cross-tab search update ships. This is the piece that closes the loop: the browser now maintains AI context across pages, across tabs, and across sessions.[Google][ALM breakdown]

Chrome is moving from “passive container for web pages” to “persistent AI research layer that sits on top of the web.” Several early write-ups also note that the enhanced AI workflow may carry meaningful memory overhead once multiple tabs and assets are in play.[ConnectQuest][Android Police]

But the performance cost is secondary to the strategic implication: the user no longer needs to leave the AI interface to get what they need. The click still happens, but the page opens inside the AI’s frame of reference. The AI remains the primary interface. The publisher page becomes evidence inside that interface.

What This Means for Brand Visibility

The Citation Economy Is Now the Primary Game

In traditional search, the goal was to rank. In AI Overviews, the goal shifted to being cited. With side-by-side AI Mode, the goal is to be cited and to survive the comparison.

The practical implications are straightforward:

  • ranking is no longer enough on its own
  • citation quality matters as much as citation frequency
  • your page now has to hold up while the AI is still mediating the interaction

When AI Mode synthesises an answer from multiple sources, it presents a curated narrative with inline citations. Research from SUSO Digital and broader AI search visibility analysis from Search Engine Land point in the same direction: only part of the citation set overlaps with traditional top-ranking organic results, and exact URL overlap can be much lower than brands expect. Ranking first in traditional SERPs no longer guarantees you appear in the AI-generated answer.

Seer Interactive’s September 2025 analysis found that when AI features are present but a brand is not cited, organic CTR drops to approximately 0.52% and paid CTR to roughly 4.14%. When the brand is cited, organic CTR recovers to approximately 1.20% and paid nearly triples to around 11.05%. The implication is clear: citation is the main mitigator of CTR decline in AI-mediated search.

The side-by-side view adds a new dimension to this. Previously, a user who clicked through from an AI answer left the AI interface entirely - that was a clean handoff. Now, the user views your page while the AI is still active and answering questions about it. If the AI can answer the user’s follow-up from your page content plus its own synthesis, the user may never scroll past the first screen of your site. If it cannot - if your page is thin, poorly structured, or contradicts the AI’s summary - the user moves on instantly.

This creates a new competitive dynamic: your page is being evaluated in real-time, in direct comparison with the AI’s own synthesis, on a split screen. You are no longer competing with the other nine organic results. You are competing with the AI itself.

Zero-Click Acceleration and the “Ghost Citation” Problem

The zero-click trend was already severe before this update. Multiple industry analyses now argue that AI answer surfaces are compressing click-through dramatically, even if the exact percentages vary by dataset and query class.[Resignal][Evergreen Media][Discovered Labs]

Two shifts matter most here:

  • the click becomes a quick validation step, not a destination visit
  • citation without brand recall creates a visibility leak

Side-by-side browsing may paradoxically increase click propensity slightly because clicking a link no longer means leaving the AI conversation. But the nature of the click changes. It becomes more of a “verification glance” than a “destination visit.” That implies shorter sessions, shallower scroll depths, and narrower ad exposure windows because the viewport is physically halved.

There is also the “ghost citation” problem. Cross-platform visibility tools and audits increasingly distinguish between citation presence and explicit brand mention, because the two do not always travel together.[Search Engine Land][ZipTie AI analytics overview][Position Digital] This means even when your content is cited, your brand name may not appear. The AI links to your page but attributes the information generically. You are providing value without gaining awareness.

For practitioners tracking brand visibility, this means monitoring citations alone is insufficient. You need to track both link presence and name mentions across AI surfaces.

How This Affects Consumer Behaviour

The Funnel Is Collapsing Into a Conversation

The traditional marketing funnel - awareness, consideration, comparison, purchase - assumes distinct stages with distinct touchpoints. AI Mode compresses these into a single conversational interface.

Consider a practical example. A user asks AI Mode: “What’s the best espresso machine under £400 for a small kitchen?” The AI synthesises reviews, specs, and pricing from multiple sources. The user clicks through to a retailer’s page, but it opens side-by-side. They ask the AI: “How does this compare to the one from [competitor]?” The AI pulls context from the open tab and its own knowledge. Then they ask: “Can I get free delivery?” If the retailer’s structured data includes shipping details, the AI answers directly. If not, the user has to scroll, and many will not.

The browse-and-compare stage that used to happen across five tabs over twenty minutes now happens inside a single AI-mediated conversation in two minutes. The user reaches a purchase decision faster, with fewer touchpoints and with the AI as the primary advisor.

For e-commerce specifically, this means the product page is no longer the start of the consideration journey. It is the validation step at the end of it. The AI has already done the comparison. The product page needs to confirm, not persuade. We explored this shift more directly in The Two-Layer Problem, where we argued that inference-based discovery and protocol-based discovery are now operating in parallel and changing what a product page is for.

Search Behaviour Is Becoming Conversational and Multi-Modal

Cross-tab search changes what queries look like. Instead of typing “best running shoes for overpronation 2026,” a user can open three product tabs, add an image of their current shoes, and ask: “Which of these would be best for someone switching from these?”

This is not a keyword query. It is a multi-modal, context-rich prompt that draws on visual input, multiple open pages, and implicit user preferences. Traditional keyword-based SEO and programmatic targeting cannot capture this intent. The query never hits a search index in a form that keyword bidding can match.

As cross-tab and multi-modal search becomes habitual, the proportion of queries that look nothing like traditional keywords will grow. Programmatic campaigns built on keyword matching will see diminishing reach - not because the audience has shrunk, but because the query surface has changed shape.[McKinsey][Pixis on SEO vs GEO vs AEO]

The Attribution Crisis

AI-mediated browsing creates a structural attribution problem. When a user discovers your brand through an AI Mode citation, views your page in a split-screen frame, and later converts through a direct visit or branded search - how do you attribute that conversion?

In practice, three failures stack on top of each other:

  • discovery happens inside the AI layer
  • traffic often arrives without clean referral context
  • conversion may happen later through an apparently unrelated direct or branded visit

Current analytics infrastructure cannot answer this reliably. Google Search Console does not distinguish AI Mode traffic from traditional search. AI browsers and agents often operate in sandboxed environments that do not load JavaScript, accept cookies, or maintain session history.[Stape][Search Atlas] Traffic from AI-mediated discovery frequently appears as “Direct” or “(not set)” in Google Analytics 4. That verification gap is exactly what pushed us to build the experimental AI Commerce Lab, where one of the core lessons was that synthetic optimization signals are much easier to generate than to validate against observed reality.

The practical consequence: marketing teams that rely on last-click or even multi-touch attribution are systematically undervaluing their AI-driven visibility. That attribution blind spot is now being discussed openly in AI search measurement work.[Discovered Labs][LSEO attribution overview]

The impact on paid search is more severe than on organic. Seer Interactive’s data shows paid CTR declining by approximately 65% for query types where AI features are present. The mechanism is straightforward: AI Mode’s synthesised answers occupy the cognitive prime position. The ad inventory that used to sit at the top of the page now competes with an AI-generated answer that the user has already accepted as the starting point.

Google is aware of this tension. At the January 2026 NRF conference, alongside the UCP announcement, Google introduced Direct Offers - a new Google Ads pilot within AI Mode that lets advertisers surface exclusive discounts when the AI detects high purchase intent.[Google Cloud on agentic commerce][Next Millennium analysis] The format is designed for conversational contexts rather than the traditional paid search layout.

This suggests Google recognises that traditional ad formats are losing efficacy in AI-mediated environments. Programmatic teams should prepare creative formats and targeting logic optimised for natural-language intent signals rather than keyword matches.

Agentic Commerce and the Universal Commerce Protocol

The AI Mode side-by-side update does not exist in isolation. It is part of a broader architectural shift toward agentic commerce where AI agents mediate purchases rather than just discovery. If you want the broader market framing for that transition, we covered it in Agentic Commerce: The $5 Trillion Shift Rewriting How Humans Shop.

Up to this point, the story is mostly about visibility and attention. From here on, it becomes a commerce infrastructure story: once the AI stays in the loop during comparison, the next question is how it completes the transaction.

What UCP Actually Is

Google launched the Universal Commerce Protocol at the NRF conference in January 2026.[Google Cloud][Google Merchant docs] The protocol was introduced with broad ecosystem participation from commerce, payments, and platform partners.

For a practitioner, the shortest useful definition is:

  • UCP is the machine-readable commerce layer
  • it gives agents a standard way to query product, checkout, and post-purchase capabilities
  • it reduces the need for one-off integrations between every merchant and every AI surface

UCP is an open standard that gives AI agents a common way to interact with merchant backends across the full commerce journey, from product discovery through checkout and post-purchase support.[Google Merchant docs] The core engineering problem it solves is what Google and implementation partners describe as the “N × N integration bottleneck”: without a standard, every retailer needs custom integrations for every AI surface. UCP collapses that into a single integration point.

The protocol is modular. Implementation guides describe UCP as deliberately layered - separating core transaction primitives (checkout session, line items, totals) from capabilities (Checkout, Orders, Catalog) and extensions (fulfillment, loyalty, subscriptions). Merchants choose which capabilities to support. Agents negotiate which capabilities they can handle.[Presta blueprint][Bluestone PIM]

In April 2026, UCP received significant capability updates: multi-item cart support, a Catalog capability for real-time product details, and identity-linking flows so shoppers can receive loyalty benefits on UCP-enabled platforms.[Google Merchant docs][Treasure Data on identity resolution]

Why This Matters for E-Commerce Teams

Chrome’s Auto Browse feature - which lets Gemini navigate sites, add items to carts, and complete reservations - is the consumer-facing expression of this infrastructure. When Auto Browse encounters a UCP-enabled merchant, the agent can execute the entire purchase flow without the user ever visiting the merchant’s website in a traditional sense.

This inverts the e-commerce model. The product page becomes an API surface rather than the main interface. Checkout becomes a protocol capability rather than a design differentiator. The competitive advantage shifts from “best shopping experience on your site” to “most complete, accurate, and machine-readable product data in the agent’s decision set.” That is the same structural inversion we described in both Agentic Commerce: The $5 Trillion Shift Rewriting How Humans Shop and The Two-Layer Problem: the merchant experience still matters, but the machine-readable layer increasingly determines whether the merchant is seen at all.

For practitioners: if your product data is incomplete, inconsistent, or not machine-readable, your products are functionally invisible to the agent. It is not that they rank lower - they are excluded from the decision set entirely because the agent cannot verify the information it needs to make a recommendation. That exclusion logic is the core warning in The Two-Layer Problem.

What Digital Marketing Practitioners Should Do Now

If you do not own the engineering layer, start with visibility and content readiness.

Immediate: AI Visibility Audit

Query your brand name, product categories, and key competitors across Google AI Mode, Gemini, ChatGPT, and Perplexity. Document whether you are cited, whether the information is accurate, and what the sentiment is. Tools like Semrush’s AI Visibility Toolkit, ZipTie, or Otterly AI can automate this, but manual spot-checking is essential to understand the qualitative experience.

Track two metrics that are not yet in most dashboards: citation frequency (the percentage of relevant AI prompts where your domain is cited) and brand mention accuracy (whether your brand name actually appears when your content is cited, or whether you are a ghost citation).

Short-Term: Restructure Content for Extraction and Comparison

AI Mode selects sources based on perceived trust, relevance, and formatting not just rankings. Content structured around clear questions with direct, concise answers (40–80 words) followed by supporting detail performs best for AI extraction.

Prioritise: question-based H2 headings that match conversational queries; short paragraphs (2–4 sentences); comparison tables with structured data; and FAQ sections that address the follow-up questions a user would ask in a side-by-side session.

The side-by-side view means your page will be evaluated while the AI is still answering questions about it. Structure content so the first screen answers the most likely follow-up. If the AI can get the answer from your page faster than the user can scroll, you win the comparison.

Medium-Term: Fortify Your Off-Site Citation Footprint

AI Mode synthesises from multiple sources. Your brand must appear across trusted platforms beyond your own site: substantive contributions to Reddit, Quora, Stack Overflow, and industry forums (non-promotional, helpful answers); accurate profiles on Crunchbase, LinkedIn, and relevant directories; guest posts on niche and mainstream publications; and earned mentions from high-authority domains.

Original research and proprietary data are particularly valuable. AI Mode heavily cites studies, surveys, and benchmarks. First-party data positions your brand as a primary source rather than something the AI paraphrases through secondary coverage.

Long-Term: Prepare for Agentic Commerce

If you operate in e-commerce, the Merchant Center is no longer just a feed upload tool - it is the primary interface between your product catalog and AI-driven discovery. Google is rolling out dozens of new data attributes designed for conversational commerce: answers to common product questions, compatible accessories, substitutes, and more.

Implement comprehensive product schema markup with gtin13/mpn for universal product matching, real-time offers.price and offers.availability synced to backend systems, shippingDetails for total cost calculation, and consistent brand naming across all feeds.

Evaluate UCP readiness. Google is simplifying UCP onboarding in Merchant Center, and the supporting merchant documentation now makes the implementation path much clearer.[Google Merchant docs][Merchant Center structured data help][Hashmeta setup guide]

What Ad-Tech Engineers Should Do Now

If you do own the engineering layer, the job is to make your data and transaction surfaces agent-readable.

Server-Side Tracking Is No Longer Optional

AI browsers and agents operate in execution contexts that break traditional client-side tracking. JavaScript may not load. Cookies may not be accepted. Session history may not persist. The result: broken attribution logic and traffic appearing as Direct or (not set).

Server-side data collection captures signals at the point of collection, normalises data across surfaces (web, app, agent), and enables more accurate identity resolution. If you have not already implemented server-side tracking via Google Tag Manager’s server-side container or an equivalent, this is now a blocking dependency for understanding AI-mediated traffic.[Stape][Cometly]

Build Agent-Ready APIs and Product Feeds

For e-commerce platforms, develop APIs with fast response times, clean JSON-LD structured data, and logical endpoint architecture for product queries, inventory checks, and pricing.

Enhance Merchant Center feeds with rich attributes that answer conversational questions - compatibility, accessories, substitutes - plus real-time inventory and pricing APIs. The new Catalog capability in UCP means agents can now query your product details in real-time. If your data is stale or incomplete, the agent will skip you and recommend from a competitor whose feed is current. This is also where the practical experimentation work from our AI Commerce Lab becomes relevant: the operational problem is not only content quality, but whether the downstream system can validate and trust the data you expose.

Prepare for Direct Checkout Integration

UCP and AP2 (Agent Payments Protocol) will require payment and fulfillment APIs that support agent-mediated transactions. The critical engineering challenge is shifting fraud prevention from “block all bots” to “enable legitimate agents while blocking malicious ones.” Agents must be authenticated. Every action must be attributable to a unique Agent ID.[Developer’s Guide to AI Agent Protocols][FinTech Weekly]

Build AI Visibility Tracking Into Your Measurement Stack

Traditional analytics are failing to capture AI-mediated journeys. Engineers should implement custom tracking for AI referral patterns, monitor citation frequency by platform using available APIs, and track “implied conversions” - branded search volume and direct traffic that follows AI exposure without a direct click.

New attribution approaches are emerging that try to assign partial credit to AI-mediated discovery by monitoring downstream effects such as branded search lift and direct traffic changes.[Discovered Labs][Search Engine Land]

Connection to the Broader Series

If you want the infrastructure underneath this shift, these are the next pieces to read:

  • The agent anatomy and orchestration guide established how role, memory, tools, and skills compose into working agents. Chrome’s new Skills library is Google’s consumer implementation of that pattern.
  • The tools, MCP, and CLI guide explained how tool connectivity becomes operational infrastructure. UCP is the commerce-specific layer that makes those machine actions transactable.
  • The empowerment imperative argued that education beats prohibition. The Chrome AI Mode rollout is proving that thesis at consumer scale: users are already experimenting, whether marketing teams are ready or not.

The infrastructure of visibility must follow the architecture of discovery. Chrome’s AI Mode has changed the architecture. The adjustments outlined above are where the practical work begins.

The Honest Assessment

A few things worth being transparent about, in line with how we approach these topics:

The data is directional, not definitive. The zero-click figures come from industry analyses with varying methodologies, not from a single canonical Google dataset.[Resignal][Evergreen Media] Google has not published official click-through data for AI Mode. The Seer Interactive CTR data is from September 2025 - before the side-by-side update. Treat these as indicative of a trend rather than precise measurements.

We do not yet know how side-by-side affects ad rendering. When a publisher page loads in a split-screen frame inside Chrome’s AI Mode, the viewport is physically halved. How this affects ad viewability, ad load, and revenue attribution is an open question. Index Exchange reported in April 2026 that 69% of publishers on its exchange experienced year-over-year ad opportunity declines throughout 2025, with an average decline of 14%. Whether side-by-side browsing accelerates or partially mitigates that trend is unknown.

UCP is early. The protocol is live and real merchants are onboarding, but the ecosystem is still forming.[Google Merchant docs] The broader agentic commerce projections - including McKinsey’s framing of AI search as the new front door to the internet - are directional, not guarantees. Build for the direction, but test before you commit significant infrastructure investment.

The window is narrowing. Google is actively testing sponsored ads within AI Mode. Brands that establish organic AI visibility now will benefit from compounding citation frequency. Those that wait will find the answer space increasingly occupied by competitors and paid placements.

Test it yourself. Measure carefully. And do not trust any single system’s recommendations - including this one - without verification.

Other Articles