Agentic Commerce Has Shipped — And Conversion Is Three Times Worse: The First Hard Data on B2C AI Checkout

See also (wiki): assistive-to-agentic-shift · ai-vendor-contracts

Executive Summary

• Forrester analyst Lily Varon’s April 9, 2026 assessment is the first public data point in the corpus on what actually happens when consumers try to buy through an AI agent: Walmart’s conversion rate inside ChatGPT Instant Checkout was roughly three times lower than when the same customer clicked through to Walmart.com.
• OpenAI pulled Instant Checkout in March 2026 — one month after launch — and pivoted to ChatGPT apps and product discovery. That is the clearest “overhyped too early” signal to date from a provider that built the protocol.
• Audience size is not the constraint. ChatGPT reached 900 million weekly active users (Feb 2026); Gemini 750 million monthly (Q4 2025); Alipay’s AI Pay cleared 120 million transactions in a single week (Feb 2026). The constraint is conversion, trust, and protocol fragmentation.
• Four incompatible protocols are now in market: Google UCP, OpenAI ACP, Alibaba ACTP, and Stripe/Tempo/Visa’s Machine Payments Protocol. No mid-market retailer can build to all four; picking one means betting on a horse with an unclear finish line.
• Stripe’s own 2026 annual letter labels agentic commerce “overhyped too early” while calling the long-run opportunity “generational.” Forrester places the industry at stage one or two of a five-stage evolution.
• The Monday-morning question for CMOs and COOs at B2C retailers is narrower than the hype implied: not “do we need an agentic commerce strategy” but “which of the four protocols, if any, do we list product feeds against this quarter, and what conversion drag do we plan for?”

The Conversion Drag Is the Headline

Every agentic-commerce piece through Q1 2026 leaned on executive-intent surveys — percentages of CMOs who believed AI intermediaries would reshape discovery, investment forecasts, strategic pronouncements. What was missing was live conversion data.

Forrester’s Varon delivers it. Inside ChatGPT’s Instant Checkout surface, Walmart’s conversion rate was roughly three times lower than on Walmart.com for the same selection. The Walmart.com baseline sat near 70% for the measured funnel; Instant Checkout came in at approximately one-third of that.

That is consistent with a pattern retailers already know from other walled-garden commerce integrations (Instagram shopping, Pinterest buyable pins, Facebook Shops) but had not seen quantified for AI chat surfaces: when the buying decision leaves your domain, trust, payment UX, returns-visibility, and loyalty-program linkage all degrade simultaneously, and conversion degrades with them.

The move that followed is the second data point. OpenAI announced Instant Checkout in February 2026, shuttered it in March 2026, and pivoted to ChatGPT apps and product-discovery surfaces. Walmart, facing the same conversion result, redirected its agentic investment from Instant Checkout to embedding its own Sparky agent inside ChatGPT and Gemini. Both moves say the same thing: the value is in discovery (AI as a top-of-funnel intent surface), not checkout (AI as a transaction surface) — at least not at the protocol maturity of April 2026.

The Protocol Fragmentation Problem

The second durable signal is that four incompatible protocols are now competing for merchant feeds:

Major processors — Adyen, PayPal, Stripe, Worldpay — are positioning themselves as connectors rather than picking a protocol, building support for multiple standards plus “Know Your Agent” identity frameworks. That is the stance a mid-market retailer should mirror: don’t bet the product catalog on a single protocol; list feeds against the two or three most likely survivors and treat the investment as optionality, not commitment.

Audience Reach Versus Adoption

The intent-side data is not the bottleneck. Three numbers make the point:

• ChatGPT: 900 million weekly active users (Feb 2026)
• Gemini: 750 million monthly active users (Q4 2025)
• Qwen: 300 million monthly active users (Dec 2025)
• Alipay AI Pay: 120 million transactions cleared in a single week (Feb 2026)

Western retailers will look at the Alipay number and assume it generalizes. It does not — yet. Alipay’s AI Pay runs inside a walled garden that already owns identity, payment, and loyalty for its users. ChatGPT and Gemini do not. Until the US and European AI assistants solve the identity/loyalty/returns gap, the Walmart conversion data is the floor, not the ceiling.

What “Overhyped Too Early” Means in Practice

Stripe’s 2026 annual letter uses the phrase “overhyped too early” while still calling the opportunity “generational.” Forrester places the industry at stage one or two of a five-stage evolution. That is not a contradiction — it is the same message the corpus hears in other AI segments: infrastructure gets built at one pace, consumer behavior shifts at another, and the first pilots that prove the gap are the most valuable intelligence a retailer can buy.

The corpus already has the executive-intent lens from IBM IBV + Adobe’s “Win the moment by mastering customer intent” (April 2026, n=1,000 senior tech and marketing executives) — the survey framing of why agentic commerce threatens the brand funnel. Forrester Varon is the companion lens: what has actually shipped, what has failed, and what the conversion drag looks like through April 2026.

Key Data Points

Source Credibility

MEDIUM–HIGH. Forrester is a paid analyst firm; Lily Varon covers commerce technology and has commercial interest in Forrester Wave reports on commerce platforms. The post is blog-form rather than a published Wave, so it sits between editorial commentary and formal research. Weighing against that: the Walmart conversion figure is the first public data point of its kind in the corpus, the timeline of launches and shutdowns is verifiable against vendor press releases, and the “overhyped too early” framing is corroborated by Stripe’s own 2026 annual letter. Apply Tier 1 freshness — published April 9, 2026.

Cross-reference: IBM IBV + Adobe “Win the moment by mastering customer intent” (April 15, 2026, n=1,000) provides the strategic/survey lens on the same phenomenon and reports $29M annual waste attributed to delay plus a 38% CLV uplift for the closed-loop personalization + governance cohort. The two pieces triangulate: IBM IBV names the strategic question, Forrester names what shipped, what broke, and where the protocol fragmentation is headed.

What This Means for Your Organization

If you run a B2C brand with direct-to-consumer revenue above $100M, the Walmart conversion figure is the number your board will ask about by the next meeting. The operational translation is narrow: the retailers that listed feeds against OpenAI ACP in late 2025 are now rebuilding against Google UCP or Alibaba ACTP, and the engineering cost of each migration is measured in integration sprints, not in strategy sessions. Your realistic posture through Q3 2026 is to treat agentic-commerce feeds as low-cost optionality — pick the two protocols most aligned with your customer base (UCP + MPP for most US retailers), list against them, and refuse to let the feature request queue consume a roadmap slot until conversion parity with your owned site is in sight.

The mistake to avoid is the opposite of the one the hype cycle warned about. The risk is not that you under-invest in agentic commerce; it is that you over-invest before the protocol winner is clear and before the conversion mechanics support the spend. Stripe’s own “overhyped too early” language from a company with every incentive to drive adoption is the most honest signal in the market right now. Treat discovery-layer AI (Sparky inside ChatGPT, Gap inside Gemini) as the higher-probability bet through 2026, and keep checkout inside your own domain until the data changes.

If your team is debating which protocol to prioritize, where to set the conversion-parity threshold that triggers deeper investment, or how to structure the merchant-feed contract so you are not locked into a protocol that gets deprecated in 12 months, I would welcome the conversation — brandon@brandonsneider.com.

Sources

1. Forrester — Lily Varon, “Agentic Payments In B2C Commerce: Where We Are Now” (April 9, 2026). Principal Analyst blog post. Credibility: MEDIUM–HIGH (analyst blog; commercial interests disclosed; first public conversion data). URL: https://www.forrester.com/blogs/agentic-payments-in-b2c-commerce-where-we-are-now/

2. Stripe — 2026 Annual Letter (referenced in Forrester piece; cited for “overhyped too early / generational potential” framing). Credibility: HIGH for stated position; MEDIUM for neutrality (Stripe is a direct participant in Machine Payments Protocol).

3. IBM IBV + Adobe — “Win the moment by mastering customer intent” (April 15, 2026, n=1,000 senior tech and marketing executives). Companion strategic-intent survey. Credibility: MEDIUM (vendor-sponsored; rigorous sample). URL: https://www.ibm.com/thought-leadership/institute-business-value/en-us/report/customer-intent

Brandon Sneider | brandon@brandonsneider.com April 2026