Where the Water Goes

How e-commerce and consumer research are mapping the same shift in AI search — and what it means for brands that compete on trust

In hydrology, when a river shifts its channel, the change is often invisible at first. The main current looks the same. But upstream, the water is already finding new paths — carving through softer ground, pooling in unexpected places. By the time the old channel starts to dry, the new one is well established.

Something similar is happening in how consumers find and choose products. The shift from traditional search to AI-powered discovery has been building for the past two years. But in the span of just a few months — from late 2025 through early 2026 — Boston Consulting Group (BCG), Gartner, McKinsey, Adobe, Salesforce, and Forrester each released findings that, taken together, suggest this is no longer an emerging trend. It’s a current that has already moved.

The scale of what’s moving

The most striking data comes from Adobe, which tracks over a trillion visits to U.S. retail sites — more than any other technology company or research firm. By mid-2025, Adobe measured a 4,700% year-over-year increase in traffic to retail sites from generative AI sources. That number alone is dramatic, but the trajectory underneath it tells the real story:

AI-driven traffic had been doubling roughly every two months since September 2024, and by the 2025 holiday season, it was up 693% year over year — with no sign of leveling off.

Gartner, approaching the question from a different angle, predicted that traditional search engine volume would drop 25% by 2026 as consumers shift to AI chatbots and virtual agents. Their consumer surveys bear this out. In a study conducted in mid-2025, 51% of U.S. consumers reported that their research habits had already changed because of generative AI. Among those who had shifted, 71% changed the way they phrase queries — using more specific terms, question-based inputs, and conversational language. Perhaps most telling: 18% now use AI tools to craft their prompts before they even open a search engine.

The old channel isn’t dry yet. But the water is clearly finding new ground.

From browsing to buying

Early skeptics of AI-powered shopping made a reasonable argument: people might browse through ChatGPT, but they’ll still buy through traditional channels. That argument has a data problem now.

Adobe’s holiday season data showed that by late 2025, visitors arriving at retail sites through generative AI sources were converting at rates 31% higher than traffic from other channels — nearly double the conversion advantage measured the prior year. Revenue per visit from AI-referred traffic climbed 254% year over year. These aren’t window shoppers. They’re arriving with clear intent and following through.

Salesforce, drawing on data from over 1.5 billion shoppers globally, found that AI influenced 20% of all retail sales during the 2025 holiday season — approximately $262 billion in revenue. Shoppers referred through AI-powered search channels like ChatGPT and Perplexity converted nine times more often than those arriving from social media.

BCG, in a January 2026 report, added that shopping-related generative AI use grew 35% in under a year, with more than 60% of consumers now expressing high trust in AI-generated recommendations.

The McKinsey and Business of Fashion State of Fashion 2026 report put a specific face on these numbers: between June and August 2025, ChatGPT accounted for 16% of Zara’s inbound website traffic. That’s not a niche early-adopter signal. That’s a major global brand seeing a significant share of its product discovery happening through a channel that barely existed eighteen months earlier.

What this means for brands that compete on trust

The research paints a clear picture for consumer brands broadly. But for brands that compete on certifications, supply chain transparency, and nuanced sustainability claims, this shift carries a specific risk that the broader research doesn’t address. It’s a risk we’ve written about before in the context of ethical brand visibility, and the new data makes it more concrete.

When a consumer asks an AI assistant to recommend sustainable running shoes, the model synthesizes information from across the web and delivers a curated answer. It doesn’t present ten blue links for the consumer to evaluate. It presents a conclusion. And the brands that appear in that conclusion — and how they’re described — depends entirely on what the AI can find, interpret, and verify about them.

This is where the shift becomes pointed. A brand with a simple value proposition (“affordable running shoes”) is relatively easy for an AI to summarize accurately. A brand with a complex story — B-Corp certification, regenerative materials sourcing, fair labor standards across a multi-tier supply chain — presents a much harder challenge. The more nuanced your differentiation, the more likely it is to be flattened or omitted entirely when an AI constructs its answer.

The Gartner data on changing query behavior makes this more tangible. Consumers aren’t just searching for “best running shoes” anymore. They’re asking questions like “running shoes made with recycled materials from companies with fair labor practices.” These are exactly the kinds of complex, multi-attribute queries where AI either gets the nuance right — citing your specific certifications and sourcing practices — or defaults to simpler, better-known alternatives.

For brands built on trust, the question is no longer whether customers are moving to AI-powered discovery. Six independent research institutions, tracking billions of transactions and surveying tens of thousands of consumers, have settled that. The question is whether your brand’s full story — the certifications, the supply chain commitments, the impact data that justifies the premium — is structured in a way that AI models can find it, interpret it accurately, and relay it with the specificity your customers expect.

The current is shifting. The brands that will thrive are the ones visible where the water is going.

---

Sources