There’s a striking asymmetry in this year’s agentic traffic data. AI agents are pouring into ecommerce sites — browsing products, comparing options, evaluating specifications — and then almost never completing a purchase. Understanding why that gap exists is the most useful thing an online store owner can do about agentic commerce right now.

What Is Agentic Commerce?

Agentic commerce is when an AI agent researches, compares, and potentially purchases products on a user’s behalf, rather than the user browsing the store directly. The agent evaluates your products against the user’s stated requirements and either shortlists you or moves on.

The commercial implication is different from traditional traffic. A human visitor can be persuaded by design, brand feeling, and copy. An agent is matching structured facts — price, specifications, availability, shipping terms, return policy — against a set of criteria. If those facts aren’t machine-readable on your product pages, you can lose the comparison before a human ever sees your store.

How Much Ecommerce Traffic Is Actually Agentic?

Ecommerce receives 38.20% of all agentic web traffic, according to HUMAN Security’s April 2026 data — second only to media at 45.62%. But only about 3.16% of agentic activity touches checkout or payment routes.

Other figures give a sense of scale and trajectory. AI traffic to US retail sites jumped 393% year over year in Q1 2026, and that traffic reportedly converts at a higher rate than average when it does convert. eMarketer forecasts AI platforms will process roughly 1.5% of US retail ecommerce in 2026 — around $20.9 billion, about four times the 2025 figure. Morgan Stanley projects agentic commerce impact could reach up to $385 billion by 2030.

Why Don’t Agents Reach Checkout?

The gap exists because checkout was designed entirely for humans. Add to cart, enter card details, click pay — that flow assumes a person with hands, eyes, and a stored payment method, and it breaks down when the buyer is software acting under delegated authority.

Several specific barriers stack up:

Payment authorisation. There’s no established, widely supported way for an agent to pay on a user’s behalf with properly scoped permission. The x402 protocol is one attempt to solve this at the protocol level, reviving the long-dormant HTTP 402 “Payment Required” status code so a server can respond with machine-readable payment terms that an agent can act on.

Identity and authorisation. Agents need a way to act on a user’s account without being handed raw credentials. Standards like RFC 9728 (OAuth Protected Resource Metadata) allow agents to send a human through a proper OAuth flow to grant scoped access.

Anti-bot defences. Many stores treat non-human traffic as hostile by default. CAPTCHAs, rate limiting, and bot-detection rules that were built to stop scrapers also stop legitimate agents acting for real customers.

Broken form mechanics. Even where an agent is permitted to proceed, checkout forms with unlabelled fields, custom JavaScript widgets, or multi-step flows that rely on visual cues are difficult to complete reliably.

What Should Store Owners Actually Do Now?

Focus on being selected rather than being transacted with. Agents are overwhelmingly in research and comparison mode in 2026, so the highest-return work is making your product data machine-readable and your store legible to agents evaluating options.

Practical priorities:

1. Get Product structured data right. Complete, accurate JSON-LD Product markup — price, currency, availability, SKU, brand, condition, shipping details, return policy — is the single highest-leverage item. This is the exact data an agent is comparing.

2. Put specifications in text, not images. Product specs embedded in an image, a PDF spec sheet, or a JavaScript-rendered tab may be invisible to an agent parsing your page.

3. Make availability and pricing unambiguous. “Contact for pricing” or availability that only appears after selecting variants creates comparison friction that pushes an agent toward a competitor with clearer data.

4. State shipping and returns in plain, structured terms. These are common shortlisting criteria, and vague policy pages are hard to evaluate against a user’s requirement like “delivers within a week, free returns.”

5. Review your bot policy deliberately. Check whether your CDN or security layer is blocking legitimate agent user agents. This is a genuine business decision, not a purely technical one — but it should be made consciously rather than inherited from a default ruleset.

6. Audit your checkout for mechanical clarity. Properly labelled form fields, standard input types, and predictable multi-step flows help agents and, not coincidentally, help humans and screen-reader users too.

Does This Change How Stores Should Think About Conversion?

It adds a stage rather than replacing one. The funnel now includes a machine evaluation step before human consideration — your store has to survive an agent’s comparison before a person ever sees it.

This makes the unglamorous parts of ecommerce more valuable relative to the visual ones. Accurate structured data, complete product specifications, and clear policy pages have always mattered somewhat; in an agent-mediated comparison they become the deciding factors, because they’re the only inputs the agent actually has.

It’s worth noting that most conventional conversion optimisation work still applies unchanged — a store that loses human customers at a clumsy checkout will lose agent-assisted ones too.

Is Agentic Commerce Worth Preparing For Now?

For most stores, yes — because the preparation work overlaps almost entirely with existing best practice. Complete structured data, clear specifications, and clean checkout mechanics improve human conversion and search visibility regardless of what happens with agentic commerce.

That overlap is what makes this low-risk. You aren’t betting on a speculative future; you’re doing the product data hygiene most stores should already have done, with an additional payoff if agent-driven purchasing scales the way current forecasts suggest.

The parts genuinely worth waiting on are the emerging payment and identity protocols — x402, agent-facing OAuth flows, and the various agentic checkout proposals. These are still stabilising, and premature integration is likely to need rework.

FAQs

Which ecommerce platforms are best prepared for agentic commerce? Established platforms like Shopify generally handle structured product data well by default, which is a meaningful head start. The bigger variable is usually how a specific store has been built and configured rather than the platform underneath.

Should I block AI shopping agents from my store? For most retailers, no — agents browsing your products represent potential demand. The judgement is different for businesses where scraped pricing data is a genuine competitive risk, but blocking should be a deliberate decision with a clear rationale.

Will agents negotiate prices or only compare them? Current agentic activity is overwhelmingly comparison and evaluation. Negotiation-capable commerce agents exist in early forms but aren’t a mainstream consideration for most stores in 2026.

How do I measure agentic traffic to my store? Check server logs and analytics for known agent user agents, and watch for sessions with human-like patterns but unusual navigation speed. Some analytics platforms have begun adding AI traffic segmentation, and Bing Webmaster Tools added AI Performance reporting in February 2026.

---

The current state of agentic commerce is a lot of looking and very little buying. That’s an unusually generous position for store owners — a window to get product data and checkout mechanics right before the transaction layer actually arrives.