Unmasking the Hidden Dangers Lurking in Agentic Commerce: What You’re Not Being Told

The agentic commerce conversation inside marketing teams narrowed to a single question: how do we get the AI to recommend us? It’s a reasonable question, but the wrong one to lead with.

A brand an agent purchases is one the algorithm selected. A brand a customer asks for by name is something a brand has to earn. As strategist Jess Graham puts it, the first case is procurement, and no one has ever fallen in love with sourcing.

That distinction matters as agents take a larger role in product discovery. A brand can be perfectly readable to the machine and completely absent from the person. The agent evaluates, the customer receives a product, and the act of choosing quietly disappears from the process.

For marketers who own martech and customer data, this is where the strategy can get tested. Algorithmic legibility matters, but so does staying present to the human being behind the agent.

AI agents are already reshaping product discovery

Agentic commerce is already in the market, though not in the shape the first headlines promised. OpenAI’s Instant Checkout went live in September 2025 and was pulled in March 2026, five months later.

Only about a dozen merchants integrated it; usage remained low, and shoppers who did their research in ChatGPT still preferred to complete the purchase on the retailer’s own site. OpenAI didn’t retreat from commerce. It moved to a discovery-first model that routes shoppers out to merchant apps and storefronts, where the retailer handles the payment.

Google went the other way with its own protocol. The Universal Commerce Protocol, announced at NRF in January 2026 with Shopify, Etsy, Wayfair, Target, Walmart, and Visa, is now live and continues to expand: cart support, catalog access, and identity linking are already in use. Read the two stories together, and one thing becomes clear. The checkout layer is still taking shape, but agents already sit in the discovery layer.

Consumer research shows AI is already influencing product discovery. Salesforce found 39% of consumers, including 54% of Gen Z, have used generative AI to discover and evaluate products. Adobe has also tracked traffic from generative AI tools to retail sites.

What the research doesn’t yet demonstrate at the same scale is agents purchasing on consumers’ behalf. Salesforce found 63% of Gen Z are interested in having AI agents make purchases for them, but that reflects stated interest rather than measured behavior.

Consider the moment this could create. A customer asks an AI agent to buy moisturizer. The agent weighs a thousand options against price, reviews, and delivery speed, and it buys one. The customer never saw the packaging, never read the brand story, and never compared it against the product they may have already used and loved. The agent made the choice, and the customer received the result.

That scene comes from Graham, who’s been mapping what happens to brands when agents start making purchasing decisions for people.

Being legible to AI isn’t enough

Graham draws a clear distinction between algorithmic legibility and brand preference.

Algorithmic legibility involves structuring product data so the agent can read it, rank it, and place a brand in the consideration set or, in her phrase, be top of algorithm. This is where budgets are going, and the early returns are real. Brands that surface inside AI-generated answers earn more traffic, and that traffic tends to spend more time on site. The work matters, and it should continue.

The other challenge is staying present to the human behind the agent. Graham’s name for the failure state is agentic invisibility. The commercial implication follows directly. A brand purchased without being deliberately chosen by a human may have surrendered its pricing power.

To an agent optimizing on price, ratings, and delivery, an undifferentiated brand reads as a commodity, and commodities compete on price until someone loses. Graham gives this a name that belongs in a P&L: the discovery tax, the compounding cost of ceding how customers find a brand and then decide it’s worth choosing again.

The cost of giving up control of discovery

Other industries have already paid the price for giving up control of discovery.

• Hotels handed discovery to online travel agencies and now pay 15% to 30% per booking while owning almost none of the guest data. 
• Musicians handed discovery to playlist algorithms and now earn fractions of a cent per stream, with the platform deciding who gets surfaced. 
• Third-party sellers built demand on a marketplace, then watched the marketplace study what sold and launch its own competing products. 

The sequence repeats in a loop. A new intermediary (in this case, AI agents) offers convenience. Brands accept worse economics to keep access. The intermediary captures the customer relationship and the data, and brand differentiation collapses into a feature comparison. 

Agentic commerce is the most complete version of this pattern so far because the customer isn’t even present when the evaluation happens.

Brand preference requires data you control

Graham’s prescription for brands is to build discovery experiences that agents can’t capture and give people a reason to choose your brand from the start. The martech stack plays a critical role because earning preference requires more than making a brand legible to an agent.

Most brands chasing legibility may be feeding agents the wrong data. Static product feeds and last-touch attribution models describe what a customer did and stay silent on why they chose. That’s enough to get ranked. Being preferred takes more.

An agent optimizing on a few structured fields treats every well-structured competitor as interchangeable. That instability shows up in AI results. 

SparkToro found less than a 1% chance that the same brand appears across two identical AI queries. A brand flickers, present in one answer and gone from the next, with no explanation offered to the marketer or the customer.

Escaping agentic invisibility then becomes a data decision before it becomes a campaign decision. It requires zero- and first-party data that captures preference and relationship signals — the ones that explain why a customer chose a brand — held somewhere inside a brand’s own martech stack.

In practice, that means owning the discovery moment rather than borrowing it (via third-party data or walled gardens, for instance). It means capturing the signals that come from community and direct conversation, treating delivery and unboxing as relationship data rather than logistics, and building consented identity that persists whether or not an agent sits in the middle.

A brand that owns the discovery, the direct relationship, and the data underneath both becomes the brand an agent can be instructed to seek by name. That’s the durable version of top of algorithm, and it runs on data the brand controls.

What marketers should do now

On the data side, the harder truth is that this is an architecture problem before it’s a strategy problem. A team that can’t connect the tools it already owns won’t suddenly capture customer preferences and signals. The agentic moment doesn’t create that gap but exposes and amplifies it while also raising the cost of leaving it open.

Start with an audit rather than a tool. Map the zero- and first-party customer data actually being captured, and be honest about it. A data set that records what was bought and almost nothing about why it was chosen is built for legibility and stops there.

Then decide, deliberately, where relationship data gets built in places that belong exclusively to your brand: owned discovery, community, the delivery moment Graham points to, and a direct channel the platform can’t freeze. Treat algorithmic legibility as the minimum owed to the machine, then invest past it.

Put a real number on the discovery tax this quarter. A board that hears “we win the transaction and lose the relationship” can fund the fix faster than one handed another dashboard.

Place the ownership of agent-facing data where the customer relationship already lives, inside marketing, rather than letting it drift into a purely technical decision made downstream.

Agentic commerce will hand the transaction to whoever is most legible to the machine. Whether it also hands over the relationship is a martech and data decision being made right now. Marketers need to lead that decision rather than leave it to default.