The Shopping Decision Is Moving Upstream

For twenty-five years, e-commerce has been built around a simple assumption: get the customer to the website, then help them decide.

Search captured the intent. Advertising generated demand. The retailer's website did the selling.

AI is beginning to break that sequence.

New data from Adobe and Similarweb suggests that discovery, comparison and evaluation are increasingly happening before customers ever reach a retailer's website. And when AI-influenced shoppers do arrive, they appear to arrive differently.

That may matter more than how much traffic AI sends.

Where is AI changing the shopping journey?

AI appears to have its greatest influence earlier in the shopping journey, particularly during product discovery and evaluation.

By upstream, we mean the parts of the shopping journey that happen before a customer reaches a retailer's website: expressing intent, discovering products, comparing alternatives and forming preferences.

According to Similarweb's January 2026 Market Research Panel, 35% of US consumers rate AI tools as most useful at the discovery stage, compared to 13.6% for traditional search engines. At the evaluation stage, AI is again rated most useful by 32.9% of consumers, compared to 15% for search. Only at the final purchase step do the two land close together.

The pattern is not that AI has replaced search. It is that a growing share of the shopping journey now happens further upstream, before the shopper reaches any retailer's website.

The shoppers who do click through from AI behave unusually well

Adobe's August 2026 AI Traffic Trends Report, drawing on more than one trillion visits to US retail sites, shows how dramatically the economics of AI-referred traffic have changed.

Twelve months ago, non-AI retail visits were worth 128% more per visit than AI visits. In July 2026, AI-referred visits were worth 53% more per visit than non-AI. That is a full reversal in a single year.

The same pattern shows across every engagement metric Adobe tracks.

Several of these advantages have persisted for a year or more, while others have strengthened materially in 2026.

AI-referred shoppers appear to arrive further along in their journey. They convert more, engage more, and leave less.

Does AI referral traffic capture AI's real influence?

Not fully. AI can shape product discovery and consideration without ever generating a direct referral to the retailer. This is the part most retailers are missing.

Similarweb's consumer research asked what people do after an AI tool mentions a brand.

Even more striking: in a separate Similarweb analysis of downstream browsing behavior, an AI-recommended brand was 2.5 times more likely on average to receive a downstream visit than the competitor it was compared against. Very little of that lift shows up as an AI referral in the retailer's analytics.

The implication is worth pausing on. A single AI recommendation can drive a Google search, a comparison shop, and eventually a direct visit to the brand's website. In the retailer's dashboard, that visit may appear as search or direct traffic. The AI recommendation that may have shaped the decision is invisible to traditional last-click attribution.

Referral traffic tells you where the click came from. It does not tell you where the decision was shaped.

Three implications for retailers

1. Visibility is not the outcome

An AI citation is an impression, not a business outcome. Useful to measure. Dangerous to confuse with the customer choosing you.

A retailer can win the AI visibility dashboard and still lose the customer. Rank in the answer without ranking in the decision. The measurement is a starting point, not a destination.

This is why AI discovery should not be treated simply as SEO for ChatGPT. SEO traditionally optimizes whether a page is found and ranked. AI systems increasingly synthesize products, evidence and recommendations into an answer. For retailers, the objective extends beyond being found to influencing whether the brand and its products are chosen.

SEO starts with being found. AI discovery increasingly comes down to being chosen.

2. Referral traffic understates influence

If a meaningful share of AI's influence on a purchase never shows up as an AI referral, traditional attribution models will underweight AI's contribution. Last-click attribution is becoming last-century attribution.

Retailers who plan the next twelve months around AI referral traffic alone risk optimizing for what is easiest to measure rather than where AI is actually shaping the journey.

3. Retailers need representation upstream

Much of the attention in AI commerce is focused on intelligence for the shopper: better assistants, shopping agents, answers, comparisons and recommendations.

But there is another side to every transaction: the retailer. As more of the shopping journey moves into AI, retailers need intelligence representing their interests on the other side.

Everyone is building intelligence for the shopper. Who represents the retailer?

What should retailers be asking?

The first step is recognizing that AI discovery is not simply another traffic channel or another SEO problem. Retailers should be asking a broader set of questions:

These are becoming commerce questions, not simply marketing questions.