AI shopping assistants are moving beyond customer-service chatbots and becoming part of the commerce interface itself. A new report from Lyro by Tidio examines evidence from Amazon, Walmart, Etsy, Salesforce and other retailers, arguing that the next challenge for ecommerce teams is not proving that shoppers interact with AI, but determining whether those interactions create incremental commercial value.

The report, AI Shopping Assistants for E-Commerce, was authored by Bart Turczynski, Head of Growth at Tidio, and draws on 39 published sources including retailer disclosures, industry research, market data and controlled experiments.

Its central argument is that retailers need to distinguish between sales associated with AI assistance and sales that occurred because of the intervention.

Amazon provides one of the largest examples cited in the report. The company said its AI shopping assistant was used by more than 300 million customers in 2025 and contributed nearly $12 billion in incremental annualized sales. Amazon also reported that customers using the assistant were 60% more likely to complete a purchase.

Those figures provide different types of evidence, however. The $12 billion figure is described by Amazon as incremental, while the purchase-completion figure represents an observed relationship between assistant usage and customer behavior. The report notes that Amazon's public earnings disclosure does not provide the experimental methodology behind its incremental-sales estimate.

That distinction is increasingly important as AI becomes embedded throughout ecommerce journeys.

An assistant can receive credit for an order even when a shopper would have purchased the same product through search, navigation or recommendations without AI assistance. In that scenario, attributed AI revenue increases without necessarily increasing total retailer revenue.

An Etsy experiment highlighted in the report illustrates the problem. A recommendation change increased recommendation clicks by 28.3%, but conversion increased by a statistically insignificant 0.22%, while gross merchandise value declined by a statistically insignificant 0.25%. Organic-search clicks simultaneously declined by a statistically significant 1.4%.

Etsy's causal mediation analysis found that the recommendation change itself produced a 0.50% conversion increase, while the resulting reduction in search activity offset 0.28 percentage points of that gain.

The findings suggest that engagement metrics can tell only part of the story. For ecommerce teams, the more meaningful measurement is whether an AI intervention creates additional purchases after accounting for behavior that would otherwise have occurred through existing channels.

That does not diminish the commercial signals emerging from retailers.

Walmart reports that shoppers using its Sparky assistant have an average order value approximately 35% higher than shoppers who do not use it. Magalu has reported conversion three times higher in its WhatsApp shopping experience than in its app. Salesforce has estimated that AI and agents influenced $262 billion in global online holiday sales in 2025, based on data covering more than 1.5 billion shoppers.

The report treats these figures as commercially significant but distinguishes them from randomized measurements of causal lift.

For smaller retailers, the immediate business case may also come from operational efficiency. UK mattress retailer MattressNextDay, a Tidio customer, reports that its assistant resolves 73% of customer conversations without human intervention and returns more than 400 hours per month to its support team.

Another Tidio customer, UK flooring retailer Renovate Direct, reports that its resolution rate increased from 42% to 69% after introducing intent-based ticket routing. Both figures are operator-reported rather than results from controlled experiments.

The report proposes proactive AI assistance as a particularly practical area for experimentation. Retailers can expose one group of comparable shoppers to an AI prompt while withholding it from another group, creating a holdout test without replacing existing search or recommendation infrastructure.

For marketers and ecommerce leaders, the implication is straightforward: AI shopping assistants should increasingly be evaluated as commerce interventions, not simply engagement tools.

Market Landscape

AI shopping is expanding across major ecommerce ecosystems, with retailers experimenting with conversational product discovery, personalized recommendations, shopping agents and AI-assisted customer service.

Amazon, Walmart and other large retailers have the resources to integrate AI across discovery and purchasing journeys. Smaller merchants face a different challenge: determining whether an assistant generates enough incremental revenue or operational savings to justify implementation costs.

This makes measurement methodology increasingly important. Click-through rates, conversation volume and AI-attributed revenue can demonstrate usage, but they do not necessarily establish causal business impact.

Strategic Outlook

The next phase of AI commerce will likely shift the conversation from “How many shoppers used the assistant?” to “What happened because the assistant was present?”

Holdout experiments provide one way to answer that question. Retailers can measure eligible sessions, exposure rates, incremental order-rate changes and average order value, then incorporate margin, returns and operating costs.

That approach turns AI shopping from an emerging technology experiment into a measurable commercial channel.

The Lyro report's broader message is that AI assistants have already become part of the ecommerce interface for major retailers. The unresolved question is how much of their reported commercial activity represents genuinely incremental value.

Top Insights

• Lyro's report argues that AI shopping assistants should be measured by incremental commercial contribution rather than attributed revenue or engagement alone.
• Amazon reports nearly $12 billion in incremental annualized sales from its shopping assistant, although the public disclosure does not detail the methodology behind the estimate.
• An Etsy experiment demonstrates how recommendation gains can be partly offset when AI-driven interactions displace existing organic-search behavior.
• Walmart, Magalu and Salesforce have reported significant commercial activity associated with AI-assisted shopping, although these measurements differ from randomized causal estimates.
• Proactive AI prompts can create practical holdout experiments, allowing retailers to compare comparable shopper groups and measure incremental order and revenue effects.

Get in touch with our MarTech Experts.