Analytics · 4 min read

74% trust AI shopping agents; only 19% use them at checkout

Salesforce research finds 74% of shoppers trust AI buying advice, yet only 19% use personal agents to check out; AI-driven Cyber Week spending is forecast at $75.8B.

By Marcus Bennett777 words

Key takeaways

  1. 74% of surveyed shoppers trust AI-agent recommendations; 20% use agents as their first research stop.
  2. Traffic from AI tools to e-commerce sites rose 104% year over year.
  3. Only 19% of personal-agent users use agents to check out; 9% grant unrestricted financial access.
  4. AI-agent users spend 8% more and purchase 23% more items per order.
  5. Salesforce forecasts $75.8B in AI-driven Cyber Week spending, up 13% from 2025.
74% of shoppers trust AI agents for buying advice – but checkout is another story
74% of shoppers trust AI agents for buying advice – but checkout is another story — Alexas_Fotos / pixabay

By Marcus Bennett

Only 19% of personal-agent users use AI to check out, even as 74% of shoppers trust recommendations from AI shopping agents, automated advisers that now handle product research and discovery. The gap defines both the near-term commercial opportunity and the operational risk for retailers.

Salesforce surveyed more than 2,500 shoppers in the US, UK, Canada and Australia. It reports that 20% of holiday shoppers worldwide use AI agents as their first stop for research and buying advice, while traffic from AI tools to e-commerce sites rose 104% year over year.

Salesforce forecasts that AI-driven shopping recommendations will drive $75.8 billion in Cyber Week spending, 13% above 2025.

The research gives no absolute traffic volume, attribution window or definition of “AI tool,” so data teams should treat the 104% increase as a directional channel signal rather than a directly comparable acquisition rate.

What does the survey establish?

The 74% trust finding covers four national markets, despite the separate worldwide framing of the holiday-shopping measure. That distinction matters when analysts compare trust, referral traffic and transaction outcomes across regions.

More shoppers encounter brands through agents: 30% said an AI agent first introduced them to a new brand, while 11% would avoid a brand if an agent told them to. Referral users were twice as likely to purchase.

Among the 39% who had used an AI agent on an e-commerce site, 80% reported a positive experience and 75% said the experience made them more loyal to the brand.

The commercial metrics are larger:

  • AI-agent users spend 8% more than the average buyer.
  • They purchase 23% more items per order.
  • Brands using AI agents record sales growth seven times faster than brands not using them.

These survey correlations do not establish that agent referrals caused the larger baskets or faster growth.

Why do agents advise rather than buy?

One in three consumers has used an AI agent. Salesforce says Meta Muse now reaches 12% of all consumers and describes Grok Bot as being in high demand, although it gives no usage figure for Grok Bot.

Among agent users, 28% assign multiple tasks per day. Common jobs include price comparisons, product reviews and brainstorming gift ideas, making brand discovery a material benefit rather than a secondary outcome.

Checkout permission remains narrow. Only 19% of personal-agent users use an agent to check out, 9% allow unrestricted access to financial details, and 36% have set up a separate credit card for agents.

The survey also finds that 73% of consumers are uncomfortable giving agents payment information, while 49% do not want agents making purchases for them. Consumers resist sharing purchase history or personal data that could lead to unwanted charges.

Reliability sets the tolerance. Shoppers will abandon an agent after two or three mistakes, and one in three consumers would blame the AI for a bad recommendation. Aggregate trust therefore does not erase concentrated session-level exposure.

A retailer can carry a strong 74% trust figure while still losing a high-value session after one consequential error. Error handling should be treated as a commercial metric, not merely a model-quality issue.

What should data teams do now?

Retailers should classify agent traffic as a distinct channel rather than blend it into ordinary referral analytics. Instrumentation should connect agent-origin sessions to product exposure, referral clicks, basket size, units per order and checkout completion.

Teams should version the data agents consume. Product feeds need current price, availability, reviews, policy and fulfillment fields so comparisons remain accurate across retailer catalogs. Recommendation confidence, inventory checks and human fallbacks should also be logged.

Payment controls require separate measurement. The 9% unrestricted-access and 36% dedicated-card findings favor bounded permissions, transaction limits, auditable approvals and explicit human review for unusual or high-value orders.

Retailers also need an operational endpoint for shopper agents, including catalog answers, policy responses and escalation paths. Salesforce argues that detecting intent, buying signals, dissatisfaction and other consumer patterns at machine scale—faster, higher-volume processing than human teams—will require investment in agentic engagement capabilities.

Salesforce expects agents eventually to make purchases with little or no human involvement, producing what it describes as a more autonomous, highly personalized and nearly frictionless shopping process. It further argues that brands investing in agentic AI will gain the most loyalty and future business.

That forecast remains directional. Salesforce disclosed the sample size and four countries, but the research summary provides no weighting details, forecast-model specification or timetable for autonomous checkout. Brands should test the 8% spending and 23% item-count effects against matched control groups before treating them as expected lift.

salesforce.com (Original)

Price tags

Share this article:

More from Marcus Bennett

Marcus Bennett

Show full bio

News editor covering people, launches and day-to-day practice in Analytics Digest is a professional daily at Analytics Digest.

10 articles

Read also

Next article »