How should brands decide which AI agents and platforms are allowed to sell their products?

By James Thomson, Managing Partner of Equity Value Advisors · Published October 12, 2026 · Last updated October 12, 2026

Short answer

Brands should apply the same discipline they use for authorized resellers. Test each agent or platform on five criteria: accurate representation, routing to authorized sellers, verified shopper authority, adequate data returned to the brand, and acceptable economics. Integrate with agents that pass, tolerate and monitor those that are unproven, and restrict those that create channel leakage or unverifiable transactions. Brands influence agents far more than they can block them.

Key points

“Allowing” an agent means choosing where to invest in feeds, protocols and terms, not flipping an on-off switch.
The Ninth Circuit’s August 2026 ruling in Amazon v. Perplexity makes legal blocking of user-directed agents harder.
A three-tier policy (integrate, monitor, restrict) gives management a clear, reviewable decision.
The strongest levers are commercial: distribution policy, authorized seller programs and data feed terms.

How much control do brands actually have over AI agents?

Less than most executives assume. An AI assistant can describe a brand’s products using public information whether or not the brand participates. An agent acting on a shopper’s instruction can visit a retailer’s site and buy. In August 2026 the Ninth Circuit overturned an injunction Amazon had won against Perplexity’s Comet agent, holding that an agent acting on a user’s instructions is the user for purposes of the Computer Fraud and Abuse Act. The case continues, but platforms and merchants now have a weaker legal tool for excluding such agents.

Policy may push further in the same direction. The AI AGENT Act (S. 5051), introduced by Senator Mark Warner on July 21, 2026, would require covered platforms to give authorized user agents nondiscriminatory access, with the FTC running a registration framework for agent providers. The bill is pending in the Senate Commerce Committee and may change.

So the realistic question is not which agents can touch the brand’s products. It is which agents the brand will actively support, which it will watch and which it will work to constrain.

What five tests should an agent or platform pass?

Representation. Does the agent describe the products accurately, using authoritative data the brand supplies or can correct?
Routing. Does it send orders to the brand or authorized sellers, or to whoever has the lowest price, including gray-market sellers?
Authority. Can it show which shopper authorized the purchase and within what limits, using verifiable credentials such as tokenized payments or signed agent identity?
Data return. Does the brand or its authorized retailer receive enough order and customer data to serve the customer and earn repeat purchase?
Economics. Are fees, promoted placement costs and pricing visibility known and acceptable, and can the brand change or exit the arrangement if terms worsen?

What does a three-tier agent policy look like?

Tier	Who belongs	Brand actions
Integrate	Agents and platforms that pass all five tests and reach meaningful shopper demand	Supply full product feeds; implement supported protocols through the commerce platform; negotiate data and fee terms; monitor monthly
Monitor	Agents that are unproven, small or pass some tests	Keep public product data accurate; test answers periodically; track referral traffic and order quality
Restrict	Agents that route to unauthorized sellers, misrepresent products or cannot verify authority	Withhold direct feeds; enforce authorized seller and MAP policies; use bot management to challenge unverified automation; escalate to platforms

Review the tiers at least twice a year. Platforms change their terms and capabilities quickly. OpenAI, for example, moved from native checkout to merchant checkout within six months of launching it.

Which levers do brands actually control?

Commercial levers

Distribution policy. Authorized seller programs, reseller agreements and MAP policies determine which offers agents can find. See channel control.
Feed terms. Direct product feeds and protocol integrations are a privilege the brand grants. Attach conditions on representation, routing and data return.
Retailer agreements. Many AI integrations run through retailers. Brands can ask retailers how they represent brand products to AI platforms and which data comes back.

Technical levers

Crawler rules. robots.txt and similar controls tell AI crawlers what they may index. Blocking crawlers usually hurts discovery more than it protects the brand.
Agent verification. Payment network programs such as Visa’s Trusted Agent Protocol, introduced in October 2025, and Mastercard Agent Pay, built on agentic tokens and announced in April 2025, aim to help merchants distinguish registered agents from malicious bots. See how a merchant knows an AI agent is authorized.
Protocols. Supporting standards such as OpenAI’s Agentic Commerce Protocol or Google’s Universal Commerce Protocol through the brand’s commerce platform reduces integration cost. Supporting a protocol does not mean accepting every platform’s commercial terms. See whether brands need to support multiple protocols.

Who should make the decision?

One executive should own agent policy, with input from ecommerce, marketing, IT, finance and legal. Without a single owner, marketing may push for every integration that adds reach while legal blocks anything that adds risk, and the company drifts into whatever platforms choose for it. See who should own agentic commerce in a consumer brand.

What good looks like

The brand has a written agent policy with three tiers, reviewed twice a year. Each integrated platform passes the five tests and has documented terms. Unauthorized sellers capture little agent-routed demand. Management can show the board which agents drive discovery and sales, and can change or exit any integration.

Frequently asked follow-up questions

Should brands block AI crawlers in robots.txt?

Usually not. Blocking AI crawlers removes authoritative brand content from the sources assistants use, which leaves resellers and outdated listings to define the products. Most brands are better served by allowing crawlers and making their own content the most accurate source.

Can a brand stop Amazon’s Buy for Me from purchasing its products on its own site?

Brands can use technical and contractual measures, and some retailers have objected to Buy for Me. The Ninth Circuit’s August 2026 ruling suggests that legal challenges to agents acting on users’ instructions face a higher bar. Brands should weigh whether the orders are profitable before trying to block them.

How many AI platforms should a brand integrate with?

Enough to avoid dependence on one, and few enough to manage well. For most mid-market brands, that means full data quality everywhere, direct integration with the two or three platforms that reach the most shoppers in the category, and monitoring of the rest.

Sources

The Next Web, Court lets Perplexity’s AI agent shop on Amazon, overturning a ban, August 5, 2026.
Davis Wright Tremaine, The Federal AI AGENT Act, July 2026.
Visa, Visa Introduces Trusted Agent Protocol, October 2025.
Mastercard, Mastercard unveils Agent Pay, April 2025.
Digital Commerce 360, OpenAI reveals updates to its agentic commerce experience for ChatGPT, March 24, 2026.

What changed in agentic commerce, and what it means for brands.

A concise weekly brief for brand executives navigating the shift from human-led shopping to agent-mediated commerce. Follow the developments that matter, the implications for brand control and governance, and the questions senior teams should be asking now.

Read The Agentic Commerce Brief on Substack