Last updated September 2026

AI is changing how people discover, research, compare and choose products online. Some of those changes are now showing up in large-scale ecommerce data, field experiments and consumer research.

This page tracks the most useful evidence on how AI is changing shopping behavior in 2026. It focuses on observed behavior rather than predictions about what AI might eventually do.

Summary

The research so far points to eight emerging changes in ecommerce.

Want to turn your website into a growth engine? Let’s chat.

1. AI traffic is growing and becoming more valuable

Traffic from AI sources to U.S. retail websites grew 393% year over year during the first quarter of 2026, according to Adobe Analytics.¹ The commercial value of that traffic is also changing. By May 2026, AI-referred retail visitors converted 54% better, spent 53% longer on retail websites and generated 53% more revenue per visit than visitors from non-AI sources.²

That is a significant reversal. A year earlier, traffic arriving from AI sources was substantially less valuable than conventional traffic. Adobe’s data covers more than one trillion visits to U.S. retail websites, although performance still varies by retail category.¹ ²

What it means for brands

AI is becoming both a discovery source and a source of higher-intent traffic. If shoppers use AI to research and narrow their options before clicking, the website may increasingly receive people who already understand the category and are closer to a decision. Ecommerce teams should measure AI referrals separately where possible and make sure product pages give informed shoppers the comparison information, proof, reviews and product detail they need to finish the decision.

2. AI has joined the messy middle

A 2026 study of 31 million Ctrip users examined how people used the travel platform’s AI assistant alongside its traditional search experience. Among journeys containing both AI chat and search, 53% involved people moving back and forth between the two. Another 26% used AI before search, while 21% searched first and then turned to AI.³

The research does not show AI replacing search. Instead, people were using each for different purposes. AI was used disproportionately for exploratory questions that were difficult to express through keywords, while conventional search remained part of evaluating actual travel options.³ (arXiv)

What it means for brands

AI appears to be joining the messy middle rather than eliminating it. This finding comes from travel, where people naturally want to inspect hotels, prices, photos and other details, so the pattern may be different for lower-consideration products. For higher-consideration ecommerce categories, brands should assume that a shopper may move between AI, search, retailer pages, reviews and brand websites before buying. The website still has a job, but that job increasingly includes helping shoppers validate or challenge a decision that has already started somewhere else.

3. AI can turn vague needs into new demand

Another Ctrip field experiment involving 15.9 million users found that making the platform’s AI assistant more prominent increased AI chat requests by 9.8%. More importantly, increased AI use led to more search queries, browsing, clicks and hotel orders while leaving click-through and conversion rates largely unchanged.⁴

The researchers found one reason for the increase. People could give AI broad requests and have the assistant translate them into specific destinations and hotel options. AI expanded searches toward more distant locations and niche leisure destinations that shoppers had not necessarily started out looking for.⁴ (Han Zhong)

What it means for brands

AI can capture demand before the shopper has decided what product, brand or even exact category they need. That makes category entry points more important. A supplement brand, for example, should not only explain the features of a magnesium product. It should establish when and why that product becomes relevant, such as sleep, muscle recovery or particular usage needs. Brands need clear associations between products and the situations, problems and needs that can cause AI to bring them into consideration.

4. Better AI can shrink the consideration set

A separate field experiment involving more than 510,000 users compared a reasoning AI assistant powered by DeepSeek-R1 with a comparable non-reasoning assistant powered by DeepSeek-V3. Access to the more deliberative reasoning assistant reduced hotel bookings by 2.5%.⁵

Users also conducted fewer searches, browsed fewer hotel options and clicked fewer hotels. The researchers found that the more informative AI answers appeared to reduce people’s perceived need to continue searching.⁵ (SSRN)

What it means for brands

Better AI can make shopping easier by giving people a satisfactory answer sooner, but that can also mean fewer products receive consideration. This study comes from travel, so the size of the effect should not be generalized across ecommerce categories. The broader issue is still important. When AI turns 50 possible choices into three recommendations, making the shortlist becomes critical. Brands need clear product positioning, complete product information and credible evidence that make it easy for AI to understand when their product belongs in that shortlist.

5. AI is beginning to create the merchandising proposition

Researchers from NYU, Georgetown and the University of Chicago analyzed more than 27,000 shopping answers produced by ChatGPT and Google AI Mode. They found that AI consistently organized recommendations using a “best for” format. One product might be presented as best for one need, another for a different feature and another for a different type of shopper.⁶

The underlying product information might contain ingredients, specifications, price, features and other attributes. The AI then decides which of those attributes to emphasize and can effectively create a new merchandising statement such as “best for sensitive skin,” “best budget option” or “best for high protein.” The researchers also found that this framing can affect the choices consumers make.⁶ (SSRN)

What it means for brands

AI is not only deciding which products enter the answer. It is increasingly influencing why each product deserves consideration. Brands should make their intended use cases, audiences, differentiators and product attributes explicit and consistent across PDPs, structured product data, retailer feeds and supporting content. Otherwise the machine helping a shopper compare the category may create a proposition for the product that the brand never intended.

6. AI may reduce some of the advantage of brand reputation

An August 2026 randomized experiment gave 350 participants 1,750 product choices across five categories, with some participants receiving access to an embedded generative AI assistant.⁷

When the better-known brand performed worse on the product attributes being compared, participants with AI assistance were significantly less likely to choose it. They reported relying more on technical specifications, less on brand names and experiencing less mental effort while making their decisions.⁷ (SSRN)

What it means for brands

Brand building still matters because people enter buying situations with existing awareness, memories and preferences. But AI can make it cheaper and easier to scrutinize an unfamiliar competitor. A strong name may therefore provide less protection when another product can demonstrate a better match to the shopper’s requirements. Brands need both mental availability and strong product evidence. Reputation can get a product considered, but the underlying product truth increasingly needs to survive comparison.

7. AI is increasing shoppers’ confidence in their decisions

Adobe surveyed more than 5,000 U.S. consumers about how they use AI when shopping online. Among consumers using AI for online shopping, 79% said they felt more confident in a purchase after using an AI assistant. Another 69% said they were less likely to return an item purchased with AI assistance.⁸ (Adobe for Business)

This does not mean consumers automatically believe everything an AI system tells them. It does show that AI is becoming more than another list of search results. For many shoppers, it is helping resolve uncertainty and making the eventual choice feel easier to justify.

What it means for brands

An AI recommendation can have more influence than a conventional search impression because it can explain why a product fits a shopper’s particular request. That makes accurate representation more important. Brands should test how major AI systems describe and recommend their products, particularly around important buying situations, because an inaccurate recommendation or omission may influence shoppers before they ever encounter the brand’s own website.

8. AI shopping agents do not evaluate products exactly like humans

A 2026 experiment randomized the order of 100 hotel listings across 5,000 AI shopping-agent sessions using four large language models.⁹

The agents examined more of the available results than human shoppers typically do. Search position still influenced which products were inspected, but the effect was weaker and inconsistent across models. Most importantly, the researchers found that the attributes displayed about each hotel mattered more than where the hotel appeared in the rankings.⁹ (arXiv)

What it means for brands

Traditional ecommerce has spent decades optimizing for where products appear because human attention drops quickly as people move down a search result. AI agents can process far more alternatives. If agentic shopping grows, complete and comparable product attributes may become more valuable relative to simply winning a higher position. This study tested hotels and simulated AI-agent shopping, so it should be treated as an early signal rather than a universal rule for ecommerce.

What these ecommerce statistics tell us

The most important change is not simply that consumers are moving from Google to ChatGPT. AI is becoming another participant in how buying decisions get made.

It can help someone articulate a vague need, introduce possible solutions, decide which attributes matter, narrow the consideration set, compare products and recommend a choice. The shopper can then continue through search, reviews, retailers, brand websites or return to AI again.

This creates an additional challenge for ecommerce brands. The traditional questions were whether people could find the product, understand it and choose it. Brands now also need to ask whether AI can understand the product, recognize the situations where it is relevant and represent it accurately when helping someone make a decision.

The evidence so far also gives little reason to conclude that ecommerce websites are disappearing. AI-referred retail traffic is growing and becoming more commercially valuable once it reaches a website. The more likely change is that AI takes on a larger part of the discovery and research process before the shopper arrives.

How the findings vary by ecommerce category

The research is still too early to claim that AI changes every ecommerce category in the same way.

Travel appears prominently in the current evidence because platforms such as Ctrip can observe the full journey from exploration through purchase. Travel is also a high-consideration category where shoppers naturally compare many options.

Other categories may behave differently. Products with detailed specifications, ingredients or measurable attributes, such as electronics, appliances, supplements and beauty products, may give AI more information to compare directly. Low-cost or routine purchases may require much less exploration, while expensive or experiential purchases may continue to send shoppers across AI, search, reviews and websites before a decision is made.

For now, ecommerce brands should treat these findings as emerging behaviors and test how strongly they apply within their own category.

Frequently asked questions

How is AI changing ecommerce in 2026?

AI is becoming part of product discovery, research, comparison and recommendation. Research in 2026 shows AI generating higher-value retail traffic, working alongside traditional search, creating new demand, narrowing consideration sets and influencing how products are compared.

Is AI replacing ecommerce websites?

Current evidence does not suggest that ecommerce websites are disappearing. AI increasingly influences the research that happens before a visit, while brand and retailer websites remain important for evaluating products, checking details, building confidence and completing purchases.

Are shoppers replacing Google and ecommerce search with AI?

Not necessarily. In a study of 31 million Ctrip users, shoppers frequently combined AI chat with traditional platform search. Among journeys using both, 53% moved back and forth between the two.³

Does AI-referred ecommerce traffic convert better?

Adobe reported that AI-referred visitors to U.S. retail websites converted 54% better than non-AI traffic in May 2026. Those visits also generated 53% more revenue per visit.²

Does AI favour established brands?

Not always. An August 2026 experiment found that AI assistance made consumers more willing to choose a less reputable brand when its product attributes better matched the comparison criteria.⁷ Brand remains influential, but AI can make competing products easier to evaluate.

How does AI decide which products to recommend?

There is no single universal formula. Different systems have access to different product information and use different models and ranking systems. Recent research does show that AI frequently organizes products around particular attributes and “best for” use cases, meaning product data and how clearly a product is differentiated can affect how it enters the recommendation.

Methodology

This page tracks ecommerce research and data published or updated through September 2026. Priority is given to large behavioral datasets, field experiments, controlled experiments and primary research focused specifically on AI and shopping behavior.

Some of the newest findings are working papers and have not yet completed peer review. Where findings come from a specific category, platform or geography, those limitations are identified rather than assuming they apply universally across ecommerce.

This page will be updated as new evidence becomes available.

Sources

1. Adobe Digital Insights, AI Traffic Report, April 2026. Adobe Analytics analyzed more than one trillion visits to U.S. retail websites and reported that AI-referred retail traffic increased 393% year over year during Q1 2026. (Adobe for Business)
Read the Adobe research

2. Adobe Digital Insights, May 2026 retail data. Adobe reported that AI-referred retail visitors converted 54% better than non-AI traffic, spent 53% longer on retail websites and generated 53% more revenue per visit. (Reuters)
Read the Reuters coverage

3. Yan, Zhong, Zhong and Zhou, “Shopping with a Platform AI Assistant: Who Adopts, When in the Journey, and What For,” March 2026. Study of 31 million Ctrip users examining how an embedded shopping AI is used alongside traditional search. (arXiv)
Read the research paper

4. Yan, Zhong, Zhong, Zhou and Mehta, “From Conversation to Consideration: AI Adoption and Consumer Demand,” 2026. Field experiment involving 15.9 million Ctrip users examining how AI usage activates demand and changes activity across the shopping funnel. (Han Zhong)
View the research summary

5. Yan, Zhong, Zhong, Zhou and Mehta, “Reasoning AI, Consumer Search, and Purchase: Evidence from an Online Platform Field Experiment,” revised June 2026. Field experiment involving more than 510,000 users comparing reasoning and non-reasoning shopping assistants. (SSRN)
Read the research paper

6. Ursu, Rao and Embrey, “The Help Me Search Era: Do We Choose ‘Better’ with AI?” August 2026. Analysis of more than 27,000 ChatGPT and Google AI Mode shopping responses followed by consumer experiments examining how AI frames product choices. (SSRN)
Read the research paper

7. Krol and Santamaria, “AI-Empowered Customers and the Erosion of Brand Power,” August 2026. Randomized experiment involving 350 participants and 1,750 incentivized product choices across five categories. (SSRN)
Read the research paper

8. Adobe Digital Insights Consumer Survey, March 2026. Survey of more than 5,000 U.S. consumers examining how AI is being used during online shopping. (Adobe for Business)
Read the Adobe research

9. Wadi and Ma, “Does Rank Still Matter? Position Bias When AI Agents Shop on Our Behalf,” August 2026. Experiment randomizing 100 hotel listings across 5,000 AI-agent sessions using four large language models. (arXiv)
Read the research paper