A blog post published this week by Veronika Höller, Head of Demand Generation at Tresorit and co-founder of Your PPC Doctor, argues that Google's announcements at I/O 2026 represent a structural break in how the internet works - not a product update. The piece, shared on LinkedIn on May 25, 2026, three days after publication, quickly drew reactions from practitioners across digital marketing.

From answer engine to decision agent

Höller's central claim is direct: Google is not updating Search. It is replacing it. According to Höller, the product unveiled at I/O 2026 is not a more capable version of the search engine that has existed for more than 25 years. It is something qualitatively different - a system that "no longer just finds information but increasingly completes tasks, prepares decisions, and takes actions on behalf of users."

That framing aligns closely with what Google itself demonstrated on May 19, 2026. At I/O 2026, Google introduced persistent background agents - autonomous systems that operate within Search without the user submitting a query. An information agent, for example, can continuously monitor apartment listings against a user's stated requirements and send a notification when a match appears. The user never has to return to Search and reformulate a request. That feature is set to launch for Google AI Pro and Ultra subscribers in the summer of 2026, with no pricing for those tiers included in the May 19 announcement.

Höller illustrates the behavioral shift with a concrete travel example. Previously, booking a hotel in Barcelona involved opening 15 tabs, comparing reviews, checking availability, reading blogs, searching Maps, and watching TikTok reviews. According to Höller, "the agent increasingly handles those steps itself." That compression of a multi-step research journey into a single delegated instruction is, in her framing, the core of what has changed.

The collapse of the click-based model

The implications for the economics of the web are not subtle. Höller describes a model where the traditional sequence - user searches, Google returns links, user clicks, website convinces, conversion happens - "became the foundation of the modern internet." That exact model, she argues, is now starting to break.

Google AI Mode has already surpassed one billion monthly active users globally, with queries running on average three times longer than conventional search queries and total volume more than doubling every quarter since launch. Those figures came from data published on May 19, 2026, alongside the I/O announcements. The scale matters: a surface used by a billion people monthly, where the model is being replaced and autonomous agents are being introduced, is not an experiment.

The traffic data published ahead of I/O reinforces Höller's structural point. Organic click-through rates for informational queries featuring AI Overviews fell 61% between June 2024 and September 2025, from 1.76% to 0.61%, according to Seer Interactive research covering 3,119 queries across 42 organisations. Paid click-through rates on those same queries dropped 68%, from 19.7% to 6.34%. SISTRIX data from March 2026 showed the click-through rate at position one falling from 27% to 11% when an AI Overview appears - a loss of nearly 60% of clicks a top-ranking page would otherwise receive. Small publishers have already lost 60% of search traffic over two years.

Höller's argument is that the arrival of agents makes this trajectory irreversible and accelerates it. When AI systems compare, filter, prioritize, summarize, book, contact, and purchase on behalf of users, she writes, "the role of websites changes automatically."

What websites become

The piece does not claim websites disappear. What it argues is that their purpose shifts. According to Höller, websites increasingly become "trust signals, data sources, entity hubs, AI verification layers" rather than "digital brochures for humans."

That formulation echoes a concern that has been building across the industry. Google AI agents were already experimenting with table reservations and task completion directly within search results as early as November 2025, when John Mueller described agentic features as allowing the AI to "navigate your website and help users with tasks there." In that framing, the website still exists - but the visitor is not human. The agent navigates on the user's behalf. Whether the website owner sees a session in Analytics depends on implementation; whether they see a conversion may depend on whether their checkout infrastructure integrates with Google's payments layer.

Google launched agentic checkout tools for the 2025 holiday season, with its product database holding more than 2 billion listings updated every hour according to Google's technical specifications. The Universal Cart announced at I/O 2026 extends that logic across Search, YouTube, Gmail, and the Gemini app - a single cross-merchant cart that checks for price drops in the background once an item is added. The user's intent has been captured; Google manages the execution.

Search as a continuous operating system

One of the more striking parts of Höller's analysis concerns the shift from reactive to continuous. Search used to wait. A user formulated a query, submitted it, and received results. That discrete, on-demand model defined what a search engine was.

Google's search chief Nick Fox outlined the technical architecture behind this shift in April 2026, describing query fan-out, agentic features, and personal intelligence in unusually granular detail. The background agents announced at I/O 2026 are the consumer-facing product of that architecture. They monitor, in the background, for information the user has expressed an interest in - without requiring the user to actively search.

According to Höller, this makes Search "designed to continuously assist," monitoring "preferences, behaviors, contexts, tasks, schedules, interests." She describes the lines between Search, Assistant, Browser, Shopping, Productivity, Commerce, Recommendations, and personal operating systems as "starting to collapse." That framing is consistent with what Google's top AI leaders described internally at I/O 2026: Liz Reid, who leads Search, alongside Jeff Dean and Koray Kavukcuoglu described a product trajectory explicitly moving away from isolated synchronous queries toward asynchronous, background workflows.

The trust question

Höller identifies trust as the mechanism AI agents will use to decide what to surface. This is, she argues, the question that will actually determine commercial outcomes in the coming years - not rankings, not click-through rates, not landing page conversion rates.

According to Höller, AI agents will favor systems that appear "consistent, trustworthy, structured, authoritative, machine-readable, contextually reliable across platforms." That shifts the relevant optimization surface substantially. Under that model, Reddit, LinkedIn, YouTube, podcasts, communities, reviews, PR, and creator content become important not because of SEO value but because "AI is not just reading websites. AI is trying to understand reputation."

Research published on PPC Land has documented that AI search traffic from platforms like ChatGPT, Copilot, and Perplexity already converts at substantially higher rates than traditional search traffic - with sign-up conversion rates of 1.66% from large language model traffic compared to 0.15% from search. That differential suggests the visitors who do arrive via AI referral are pre-qualified in ways traditional clicks are not, which is one early data point in favor of Höller's trust-as-ranking-signal thesis.

Google has been training its deep search agents using a dual-agent architecture - a data generator agent creates question-answer pairs by iteratively searching and gathering information, while a separate search agent attempts to solve the generated questions. The training methodology itself reflects how the system prioritizes structured, authoritative sources capable of supporting multi-step reasoning chains.

The open web's stress test

Höller is direct about the structural stakes. "Today's internet economy depends on publishers creating content, users clicking, traffic being monetized," she writes. "But if AI delivers the answer directly - or even completes the action itself - the open web potentially loses its most important mechanism."

She argues the real question is not whether SEO will survive but "how does the internet function when people visit fewer and fewer websites?" That framing has been a recurring concern in the publisher community throughout 2024 and 2025. Media leaders surveyed by the Reuters Institute described Google Web Search traffic falling from 51% to 27% of referrals between 2023 and the fourth quarter of 2025. News publishers have seen Google search traffic drop by half in two years, with the steepest declines concentrated in informational verticals like health and education.

A PPC Land analysis from April 2026 found that 69% of publishers on the Index Exchange platform experienced year-over-year declines in ad opportunities throughout 2025, with an average decline of 14% across that group. These were slot requests - the closest proxy for monetisable traffic - not direct page views. The mechanism Höller describes - AI completing the action before the user clicks - is already producing measurable signals in ad inventory data.

The behavioral dimension

Höller's piece is notable for framing the shift as psychological rather than purely technical. "Most AI discussions focus on technology," she writes. "But the real transformation is behavioral." Her argument is that once users internalise that AI can prepare decisions, save time, filter recommendations, and handle tasks, search behavior itself changes. People will "compare less, research less, click less, manually evaluate less."

That behavioral shift is harder to measure than CTR data. But it matters for anyone building digital marketing infrastructure. If the baseline assumption - that users want to evaluate options themselves - changes, the persuasion architecture built on top of that assumption (landing pages, A/B testing, conversion funnels) may be optimizing for a user who is increasingly rare.

Google's own internal sessions at I/O 2026 included a warning from a Google engineer that AI will multiply software output by an order of magnitude or more, and that systems built to handle today's volumes will not hold. That warning was directed at developer infrastructure, but it applies equally to marketing measurement systems built for human-paced interaction.

The question Höller poses

The post closes with a formulation that has already circulated widely in the marketing community. According to Höller, the future question is no longer "How do I rank on Google?" but "Does AI understand and trust my business at all?"

Sharon Peredo Sahid, a Global Digital Marketing and Product Strategist who commented on the LinkedIn post, wrote: "I applaud this, support it and declare myself loyal to this approach. Very nicely described."

That reaction reflects how the piece has landed among practitioners who have been watching the traffic data and the I/O announcements simultaneously. The search engine that underpinned digital advertising for more than two decades is being rebuilt into something that operates on a different logic entirely. The numbers in the traffic research and the architecture in Google's technical sessions are consistent with Höller's framing. How long the transition takes - and what digital marketing looks like on the other side - remains unresolved.

Timeline

Summary

Who: Veronika Höller, Head of Demand Generation at Tresorit and co-founder of Your PPC Doctor, writing on her personal blog and LinkedIn.

What: A detailed analytical piece arguing that Google's announcements at I/O 2026 represent a structural transformation of Search from a passive link-retrieval engine into an active decision and task-completion agent, with significant consequences for the economics of digital marketing and the open web.

When: The article was published on Höller's blog and shared on LinkedIn on May 25, 2026, three days before this report; the I/O 2026 announcements it analyses were made on May 19, 2026.

Where: Published on Höller's personal blog at veronika-hoeller.de and shared via her LinkedIn profile. The events analysed took place at Google I/O 2026, livestreamed from Google's campus, and are documented across multiple PPC Land articles.

Why: The piece responds to a gap Höller identifies between the scale of the structural change announced at I/O 2026 and the conversation still happening in the marketing industry, which according to her remains focused on AI Overviews, rankings, and CPCs while a more fundamental shift - from search engine to digital agent - is already underway.

Share this article
The link has been copied!