Why the Smartest Brands Are Already Rebuilding for a World Where Google Stopped Sending Them Traffic
The search-driven commerce model that underpinned two decades of digital retail is finished, and the brands that have grasped this are no longer optimising for it. At Google I/O 2026 a fortnight ago, the company confirmed that AI Search has become the global default, that AI Mode has crossed 1 billion monthly active users, and that its new Universal Cart now compresses the entire journey from query to checkout into a single surface Google owns. For any business that built its growth on ranking in search results and converting the resulting click on its own storefront, the foundation has gone.
The immediate consequence is a collapse in the traffic that fed that model. Organic search referrals have fallen by roughly a third since AI Overviews became the default, because the answer the user wanted now arrives inside Google rather than on the page the user once clicked through to reach. The deeper consequence is structural and harder to reverse. When Google can resolve a query, compare the options, and complete the purchase without the user ever leaving its interface, the brand at the end of that journey is no longer a destination. It is a supplier to a platform that now stands between it and its own customer.
Mohammed AlThaher, Regional Vice President for MEA at Sitecore, has been telling the company's clients to stop defending a position that no longer exists. "The traditional model, where a brand optimises for keywords to drive traffic to a destination storefront, is being redefined — becoming obsolete," he said. "As Google shifts to AI Overviews, the path from query to cart is being compressed. Brands need to shift their thinking away from ranking and optimise for AI visibility. In short, they have to ensure their content is structured and authoritative enough to be surfaced by AI algorithms as the best answer, rather than a link on a list."
The discovery layer did not weaken, it changed owner
To understand why the commerce model broke, it helps to understand what changed in the act of searching itself. For 25 years, Search was a retrieval mechanism that matched words to pages and returned a ranked list, and the entire economics of digital retail flowed from the outbound click that list generated. The Intelligent Search Box unveiled at I/O, described by Google as the most significant update to the interface in a quarter of a century, no longer behaves that way. It accepts video, PDFs, audio and open browser tabs, it sustains a conversational thread rather than returning a static block of results, and its agents can act on a request rather than merely answer it. Liz Reid, Google's VP of Search, summarised the new function in a sentence that ought to alarm every retailer who heard it, saying that whether a user wants to find something, buy it or book it, Search will help them get it done.
Finding was always the purpose of Search, while buying and booking were completions that happened elsewhere, on websites owned by other companies and reached through the click Google's advertising business was built around sending. The Universal Cart removes that final step from the open web. Andreas Hassellöf, Chief Executive of Ombori, has watched the product drift toward this point for some time and describes the shift in terms of what the user now experiences. "Google is becoming more like ChatGPT and Claude instead of a pure search engine," he said. "The convenience is real. People can now ask longer, more natural questions and get synthesised answers that feel helpful right away."
That convenience is exactly what relocates power away from the brand. When the purchasing decision begins with a synthesised answer rather than a list the user evaluates for themselves, the answer becomes the point of sale, and the brand's fate depends on whether the model surfaces it. Hassellöf is direct about how unreliable that arbiter can be. "Yet most people still do not understand how an LLM actually works," he said. "They treat the first confident response as fact. Hallucinations keep showing up. Google's own AI Overviews have suggested adding glue to pizza sauce to make cheese stick better. They have invented idioms such as 'you cannot lick a badger twice' and described them as real sayings. They have cited scientific papers that do not exist and given wrong details on everything from history to video games."
When the answer can be bought, the brand cannot trust it to be neutral
For a retailer, the reliability of the model is not an abstract concern about misinformation. It is a direct question about whether the system that now decides which products a customer sees can be trusted to decide on the merits. Hassellöf draws a distinction that matters here, between the obvious errors Google has largely trained out and the plausible ones it has not. "Obvious hallucinations like these are becoming rarer because companies spend extra time shaping and training the models to avoid clear mistakes," he said. "But non-obvious hallucinations are still very common. These are the plausible but incorrect claims the model presents as solid fact. When this becomes the default way billions of people get information, the risks to society grow. Misinformation spreads more easily and carries more weight because users accept the first confident answer without checking."
Layer commercial incentive on top of that opacity and the position for brands becomes plain. The old search results page was imperfect, but its paid placements were labelled, however imperfectly. A synthesised recommendation carries no such marker, and Hassellöf warns that the line between answer and advertisement could disappear entirely. "If Google now brings the same commercial logic to its new AI-generated responses, the risks grow even more," he said. "Hallucinations could blend with sponsored content or privileged partnerships. Money becomes the gatekeeper instead of the most trusted source. Users might accept commercially influenced answers as neutral truth without realising it." For a brand, the implication is unambiguous. Visibility inside Google's answer is no longer something that can be earned purely on quality, and it cannot be controlled from the outside at all.
What brands can still own when the transaction belongs to Google
This is the point at which the better operators have stopped lamenting the loss of the old funnel and started rebuilding around what remains theirs. AlThaher's argument to Sitecore's clients is that the answer is not to fight Google for the transaction, which is a contest brands will lose, but to own the layers the transaction cannot reach. The first of those is the direct relationship with the customer and the data it produces. "In the era of agentic commerce, purchasing power will shift toward AI agents and platforms," he said. "We may reach a point where the human and the brand only engage post-purchase."
If the engagement that remains is post-purchase, then the brand experience and the data underneath it become the entire competitive field. AlThaher's prescription is to consolidate rather than fragment. "Brands need to deploy their own AI-enabled digital experience platforms to manage directly the data interaction with the AI agents," he said. "The brand experience also needs to be stepped up, with brands uniting their content, data and commerce into a single ecosystem so there's a consistent message across each and every digital channel. It is also important to deliver omnichannel personalisation, given each customer's preferences. In the era of AI, this isn't just possible; it's mandatory."
The asset that ultimately determines whether a brand survives the transition is the one Google cannot replicate, namely the proprietary record of who its customers are and what they want. A retailer that holds genuine first-party behavioural data can negotiate with the platform layer from a position of knowledge rather than dependence, and can sustain a relationship that exists whether or not Google chooses to surface it. "Retailers should utilise consumer data platforms to collect unique behavioural, preference, and loyalty data," AlThaher said. "This allows brands to bypass platform gatekeepers by nurturing distinct one-to-one relationships with each and every one of their customers."
The window is closing faster than the strategy decks suggest
The danger for most businesses is not that they fail to understand the shift but that they treat it as a problem for the next planning cycle rather than the present one. The infrastructure is already deployed, the billion users are already inside it, and the brands building their own data and experience layers now are doing so while there is still a relationship left to capture. Those waiting for the picture to clarify will find that the customers whose loyalty they hoped to win have already been intermediated by an agent acting on the customer's behalf and, increasingly, on the platform's terms.
None of this is settled in Google's favour by default, and it would be a mistake to read the consolidation as either inevitable or benign. Regulators in several jurisdictions are examining what it means for a single company to own discovery, transaction and the verification of truth at once, and the same opacity that worries Hassellöf about misinformation worries competition authorities about self-preferencing.
For the brands themselves, the cost of the transition is not only strategic but human, falling on the marketing teams whose search expertise is being devalued overnight and the smaller retailers without the resources to build a data platform of their own.
The businesses that will come through this are not the ones that ranked best, but the ones that understood early that the only thing Google could not take from them was the direct relationship with the customer, and built to protect it before the architecture hardened around them.
