This is a sourced chronology of the agentic-commerce events that shape how AI shopping agents select stores, products, and offers: protocol launches, market shifts, and the research we cite across this handbook. Each entry is typed by evidence, sits under the year it happened, and links its primary source. The list is newest first, it is not exhaustive, and it grows as verified events land. For the running log of smaller changes to this site and to how agents choose, see the updates log.
2026
Adobe reports retail AI-driven traffic up 393% year over year in Q1 2026
Agent-referred demand is already arriving, and in March 2026 it converted 42% better, so being selectable by agents is a present concern, not a future one.
Adobe finds retail product pages average 66 out of 100 for machine readability
Roughly a third of a typical product page is not machine-readable to an agent, so closing that gap is a concrete advantage most catalogs have not yet taken.
Google introduces the Universal Commerce Protocol (UCP) at NRF 2026
UCP proposes one open standard spanning discovery, checkout, and post-purchase, with the retailer remaining the merchant of record.
Source: Google (blog.google): Sundar Pichai's NRF 2026 remarks (opens in new tab)
2025
Visa and Cloudflare announce the Trusted Agent Protocol
Trusted Agent Protocol lets a merchant verify that an AI agent is legitimate, an identity rail that gates whether an agent's purchase is accepted at all.
Source: Visa: Trusted Agent Protocol (investor.visa.com) (opens in new tab)
OpenAI and Stripe launch the Agentic Commerce Protocol (ACP) and Instant Checkout
ACP is the open checkout standard behind ChatGPT Instant Checkout, one of the first mainstream agent-completed purchase flows.
Source: Stripe Newsroom: Stripe and OpenAI bring Instant Checkout to ChatGPT (opens in new tab)
Google announces the Agent Payments Protocol (AP2)
AP2 defines the payment-authorization and trust layer, with signed mandates, that agent checkouts build on.
Source: Google Cloud: Announcing the Agent Payments Protocol (AP2) (opens in new tab)
ACES agent-shopping study first appears on arXiv
The study documents the selection levers this handbook tracks: grid-position bias, a sponsored-tag penalty, an endorsement lift, per-model price sensitivity, and seller-side description rewrites.
Source: What Is Your AI Agent Buying? (ACES, arXiv:2508.02630) (opens in new tab)
Mastercard and Microsoft introduce Mastercard Agent Pay
Agent Pay binds a tokenized card credential to a specific agent, merchant, and consent policy, part of the payment-rail layer a store needs to complete an agent's purchase.
Source: Mastercard newsroom (reported, corroborated by PYMNTS) (opens in new tab)
2024
Jeremy Howard proposes the llms.txt standard
The llms.txt format gives sites a root-level Markdown index for language models; its only prioritization levers are link order and an Optional section, which shapes what agents read first.
Source: llms.txt proposal (llmstxt.org, Jeremy Howard) (opens in new tab)