AI agents are learning to shop.

Shopping agents can now research products, compare options and, in some cases, complete a purchase for the customer. Adobe forecasts that traffic from AI sources to U.S. retail sites will grow 130% year over year this holiday season. The retailers that decide where those agents can go are not taking the same approach, and that matters for every brand that sells through them.

Three retailers, three sets of rules

Amazon is keeping outside agents out. Since September 20, shoppers using Meta's new Muse agent on Amazon.com have seen a message that "continued access by an unauthorized AI agent" violates Amazon's Conditions of Use, GeekWire reported. Amazon has also moved to block shopping agents from Google and OpenAI while building its own: Alexa for Shopping, launched in May, and Sponsored Products and Sponsored Brands prompts, a paid ad format that runs inside it.

Other retailers are taking different paths. Target permits agents from an approved list of companies, and Walmart has worked with OpenAI and Google to bring in shoppers through their AI tools, according to The Wall Street Journal's reporting as summarized by PYMNTS. Walmart's U.S. CEO said those efforts are bringing in new customers, describing the conversion rates as "fantastic."

Why the agent becomes the shelf

When a person shops, a brand has many chances to make its case: the search ad, the product page, the photos and the packaging. When an agent shops, much of that can be skipped. The agent reads product data, reviews and third-party coverage, narrows the options and may present only a shortlist. The customer may never see the brand's ads or product pages at all.

It is still early. BrightEdge estimates that purchases through agents are rising about 43% a month, yet only about 3% of the ecommerce companies it works with let agents complete transactions, PYMNTS reported. That gap gives brands time to get ready before agent shopping becomes routine.

Four things to consider

1. Make product data complete and consistent

Titles, specifications, pricing and availability should match everywhere an agent might look, from your own site to each retailer's listing. Gaps and contradictions give an agent a reason to choose a competitor whose information is easier to trust.

2. Read your reputation the way an agent would

Ratings, reviews and third-party coverage are the evidence an agent weighs. Ask what it would conclude about your product from that record, and address the patterns in customer feedback, not only the number of reviews.

3. Map each retailer's rules

Know which retailers allow, limit or block outside agents, and which sell ads inside their own. Amazon says existing sponsored ads campaigns are automatically eligible to run as prompts in Alexa for Shopping. Your retail media plan should reflect where agents can and can't reach, and what you are already buying.

4. Measure AI-referred traffic on its own

Set up analytics so traffic and sales from AI sources show up separately from other channels. Merchants have limited visibility into who is behind an agent's purchase, so start with what you can track today, and agree on how you will judge whether agents bring new customers or shift existing ones.

Getting on the shortlist

No one knows yet which retailer's approach will win, and the rules may keep changing. The foundation won't: accurate product information, a reputation that holds up and measurement you control. We would start there now, so your brand is ready for whichever agents your customers choose.

Questions to ask your team

• Is our product data consistent on our site and at every retailer?

• What would an AI agent conclude from our reviews and coverage?

• Which of our retailers allow, block or sell ads inside AI agents?

• Can we see AI-referred traffic and sales on their own?