Twelve months ago, “agentic commerce” was a category most merchants hadn’t heard of. Now it’s the line item that determines whether your 2026 plan is defensive or offensive.
The reason isn’t any single platform release. It’s the cumulative effect of a dozen of them. Every major LLM company shipped a meaningful commerce-adjacent capability this year, every major commerce platform shipped some kind of agent compatibility, and the buyer-behavior signal moved from “weird edge case” to “obvious enough to brief a board about.”
Here’s the 2025 ledger. Eight events that mattered, what they shifted, and what they mean for whoever runs your e-commerce team next year.
January: ChatGPT Search rolls out to all consumer accounts
OpenAI had GA’d search to ChatGPT Plus users in late 2024. In January it opened to all free-tier users. The number of consumers with a working AI search alternative to Google effectively quintupled overnight. For commercial queries with comparison-shaped intent, the shift in query share was immediate.
The merchant implication: traditional SEO budgets started spending against a surface that was shrinking, at least for the highest-intent slice of commercial queries. Most teams didn’t notice until summer, by which point the agent-mediated category was eight months into compounding.
March: Anthropic ships a production-ready Computer Use
The first version of Claude’s Computer Use shipped in October 2024 as a research preview. The March release made it reliable enough for production agent workflows. Within weeks, the first generation of consumer-facing agents that could navigate websites, fill forms, and complete purchases on behalf of users were in market.
This was the first time a buyer could meaningfully delegate the entire research-and-purchase workflow to an AI without the agent needing custom integrations with each store. The implication: any merchant whose site was hostile to a vision-and-DOM-based CUA (heavy modals, captcha walls, broken keyboard navigation, anti-bot detection) became invisible to a growing slice of high-intent buyers.
Spring: Google AI Overviews expands to all commercial SERPs
What had been a slow rollout became a default. Above-the-fold AI Overviews appeared on the vast majority of commercial search queries. Click-through rates to organic results dropped meaningfully across categories per multiple industry analyses. The position-1 organic listing that used to win the category became position-1 below the AI summary, a structurally different result.
Merchants who had built their acquisition on Google organic SEO saw real revenue impact. Many treated it as a Q2 anomaly. By Q4, it was clearly the new baseline.
Summer: Apple Intelligence ships shopping in Siri
Apple’s “shopping” surface in Siri didn’t get the press the OpenAI releases did, but it shipped meaningful capability to hundreds of millions of iPhones. Users could ask Siri to find products, compare them, and complete purchases through Apple’s payment infrastructure. The integrations were limited at launch but growing fast.
The implication is platform-specific: brands with strong Apple Pay integration and clean product schema started showing up in Siri product responses. Brands without either didn’t. Apple’s commerce surface is opt-in for merchants in a way Google’s isn’t, which means the merchants who showed up early have an asymmetric foothold.
August: Perplexity Shopping launches with native checkout
Perplexity had been a research surface. In August it became a commerce surface, with native checkout for participating merchants and a curated set of integrations. The conversion paths were narrow but the buyer-to-purchase distance was shorter than any other agent product to date.
Perplexity’s bet is that high-intent research queries naturally convert when the answer can drive directly to a purchase. Early signals in the categories where Perplexity has integrations are consistent with the bet, with Perplexity-originated traffic converting at higher rates than typical search-originated traffic. The category list will get longer in 2026.
September: OpenAI and Stripe announce the Agentic Commerce Protocol
This was the inflection point. ACP defined a specification for how AI agents could complete purchases on behalf of users without requiring custom integrations per merchant. Shopify was a launch partner. BigCommerce, Salesforce Commerce, and a half-dozen headless platforms announced integrations within weeks.
The strategic significance is that agent-mediated checkout went from a research project to a standard. The first major platforms to ship ACP-compliant integrations get the early-bird advantage in a category that, by the next holiday, will be a default expectation. The merchants who haven’t enabled their platform’s ACP toggle by Q1 2026 will find themselves competing against merchants who have.
October: The content-licensing wars compound
Reddit, Stack Overflow, the New York Times, and a dozen other content properties either struck high-priced licensing deals with LLM companies or sued them. The strategic situation crystallized. High-quality training and retrieval content has real, recurring value to the LLM companies, and the companies that own it have leverage they didn’t have a year ago.
The relevance to e-commerce is indirect but real. The agents fetching content at query time treat licensed sources differently from open-web sources. Brands whose content is high-quality and clearly licensed (or clearly free to quote, depending on the negotiation posture) get cited more confidently. The blanket “everything we publish is fair game” stance of the open web is gradually getting renegotiated.
December: ChatGPT Instant Checkout enters early access
A handful of large Shopify merchants are now testing ChatGPT Instant Checkout in production. The buyer experience is straightforward. A buyer asks ChatGPT about a product, the agent presents options, the buyer picks one, and the purchase completes inside the conversation. The merchant’s storefront is bypassed entirely for the purchase step.
The full rollout will happen in early 2026. The merchants in the early-access cohort are getting six to twelve weeks of head start. By the time Instant Checkout is GA, the leaderboard for “what catalogs agents recommend with confidence” will already have winners and losers.
What changed underneath
The pattern across all eight events: AI platforms moved from being a traffic source to being an intermediary. The buyer used to come to your site. Increasingly, the agent comes to your site on the buyer’s behalf, or doesn’t come at all because it has enough information to recommend a purchase without visiting. The conversion no longer happens on your domain.
For most merchants, the implication is structural. The work that used to win you customers (page speed, conversion rate optimization, retargeting, cart abandonment recovery) applies only to the share of buyers who still go through the human-browsing surface. The work that wins you customers on the new surface is different in kind: structured data, honest catalogs, explicit policy disclosure, protocol-compliant checkout, instrumented agent traffic so you can see what’s working.
The merchants who learned this in 2025 are the merchants who’ll outgrow the category leaders in 2026. The merchants who treat it as next year’s problem are going to spend 2026 catching up.
Next post: our 2026 predictions piece. Five things we think will be true in twelve months, with our reasoning attached. We’ll be honest about the calls we miss when we do the mid-year retrospective in June.