---
id: "claim-autonomous-checkout-difficulty"
type: "claim"
source_timestamps: ["§ Meet Your New Customer: The AI Agent"]
tags: ["conversion-rates", "product-failure", "checkout-flows"]
related: ["concept-human-present-mode", "entity-openai"]
speakers: ["Kartik Hosanagar"]
confidence: "high"
testable: true
enrichment_status: "directionally accurate; specific numbers/dates quantitatively unverified"
sources: ["geo"]
sourceVaultSlug: "hbr-seg-geo"
originDay: 3
articleStem: "hbr-tier2-05-market-to-ai-customer"
sourceUrl: "https://hbr.org/2026/06/how-do-you-market-to-an-ai-customer"
sourceTitle: "How Do You Market to an AI Customer?"
---
# Fully autonomous checkout is harder to build than anticipated

**Claim:** Fully autonomous AI purchasing is technically and behaviorally difficult to implement successfully.

**Evidence cited by [[entity-kartik-hosanagar]]:** [[entity-openai-d5]] launched **Instant Checkout** in **September 2025**, letting users complete purchases directly inside ChatGPT. It produced conversion rates **three times lower** than when users clicked through to the retailer's own site, and OpenAI **killed the feature five months later (March 2026)**. The implication: near-term agentic commerce will lean on [[concept-human-present-mode|human-present]] conditional approvals rather than fully autonomous, black-box purchasing. This also underwrites [[claim-checkout-belongs-to-retailer]].

**Confidence:** High (as stated). **Testable:** Yes.

*Enrichment status — treat as directionally accurate but quantitatively unverified.*
- **Supported:** Instant Checkout existed (OpenAI + Stripe tied ACP to it ~Sept 2025); OpenAI materially shifted *away* from AI-owned checkout toward **merchant-controlled** models; Checkout.com states "the merchant owns the checkout, not the AI platform."
- **Not publicly supported:** the exact **3× conversion delta** and the **March 2026 kill date** appear to be internal or inferred figures, not publicly verifiable metrics.
- **Downstream guidance:** do not repeat the specific numbers/dates as fact without qualification.
- **Counter-perspective:** in low-risk, high-frequency categories (household staples, digital goods) with predefined constraints, fully autonomous checkout may advance faster than the article implies as agent-authorization and fraud controls mature (see [[question-google-in-chat-checkout]]).


## Related across articles
- [[claim-checkout-belongs-to-retailer]]
- [[claim-openai-ranks-by-checkout]]
- [[question-google-in-chat-checkout]]