Shopping is splitting in two.
On one side, the customer journey we know: humans browsing, comparing, clicking, abandoning, returning. On the other side, a new surface where buyers describe what they want to an AI agent and the agent goes to fetch it.
Most brands are still optimizing only for the first one.
That’s the thing we want to plant a flag on with this blog. We’re at the start of a quiet but irreversible split in how products get bought, and the gap between the merchants who internalize it now and the ones who wait for it to become obvious is going to define the next two years of e-commerce.
What changed
In late September, OpenAI and Stripe announced the Agentic Commerce Protocol, a specification for letting AI agents complete real purchases on behalf of users. Six weeks later, ChatGPT Search is GA across all consumer accounts, Anthropic’s Claude can navigate websites and take actions through the Computer Use API, and Perplexity has rolled Shopping features into its product. Google’s AI Overviews now dominate the top of every commercial SERP. Apple Intelligence quietly built shopping into Siri this summer.
If you’re a merchant, the relevant question isn’t whether buyers will send agents to shop for them. They already are. The relevant question is what your storefront looks like to a buyer who never sees it.
The blind spot
Here’s what most brands miss: the agent isn’t a new traffic source. It’s a new intermediary. The buyer briefs the agent. The agent does the research, shortlists the options, compares the specs, sometimes completes the purchase. By the time a referral lands on your site, if it lands on your site at all, the meaningful evaluation has already happened.
Your conversion rate optimization, your A/B testing apparatus, your retargeting funnel: all designed for buyers who at some point looked at your page. When a buyer never looks at your page, those tools don’t fire.
What does fire is the agent reading your structured data, your llms.txt, your product feed, your meta description, your AggregateRating JSON-LD, the consistency between your prose and your structured claims. The agent makes a decision based on that information. The buyer reads the agent’s summary. The agent buys, or it doesn’t.
This isn’t a future scenario. We’ve watched the pattern form over the last six months. The window where “AI traffic” was a rounding error has closed.
What we believe
Here’s the thesis we’re betting on.
The agent surface will overtake search for high-intent product queries within 24 months. “Best wireless earbuds for running” is a query Google has won for a decade. ChatGPT and Perplexity are now structurally better-suited to answer it, because the answer is a comparison and a recommendation, not a list of links. The queries leaving Google first are the highest-intent ones, which is exactly the cohort merchants can least afford to lose.
Agentic checkout will become default for mid-sized baskets. Friction is the only thing keeping it from being default already. The ACP announcement is the friction-removal step. Once a buyer can complete a purchase from inside ChatGPT without leaving the conversation, they will, and they’ll get used to it.
“Agent SEO” is its own discipline. It overlaps with traditional SEO but doesn’t reduce to it. The signals are different. The ranking dynamics are different. The honesty bar is different. Agents penalize unreliable catalogs and contradicted claims in ways search engines mostly don’t. Brands hoping their existing SEO playbook covers agentic discovery are going to find out it doesn’t.
Walled-garden agent commerce and open-protocol agent commerce will diverge. Amazon and Walmart are building their own agent surfaces optimized for their own catalogs. ACP and llms.txt are the open path. The merchants who depend on third-party marketplaces will find their margins compressed. The merchants who build for the open protocols will keep their unit economics.
The accumulated learning advantage compounds. Every agent interaction with your store is a signal: what query produced the visit, what content the agent extracted, whether the citation led to a referral, whether the referral converted. The merchants who instrument this from the start build a moat against the merchants who don’t. By the time the slow-movers wake up, the leaders will have a year of training data on what works.
Why we built Inception Agents
We started this company because we noticed something specific. There was no infrastructure for merchants to be intentionally legible to AI agents.
Every major commerce platform (Shopify, BigCommerce, WooCommerce, Salesforce Commerce) was built assuming a human visitor with a browser. The signals it emits (HTML, CSS, JavaScript, product feeds optimized for Google Shopping) are designed for that visitor. The signals agents need (structured data with policy disclosure, llms.txt with per-product context, ACP manifests, JSON-LD that mirrors prose, honesty audits, real-time stock and price truth) are mostly absent or wrong.
We close that gap. We sit between your store and every major agent, detect agent traffic at the edge, serve the structured representations they need, log every interaction as a learning signal, and feed that intelligence back into what we serve next time.
The product is the closed loop. Detection, intent classification, optimized response, outcome, updated content, repeat. Each interaction makes the next interaction work better.
What this blog is for
We’re going to write about this twice a week. Not “AI is going to change everything” essays. That’s lazy. Specific posts about:
- What the major LLM platforms are shipping that affects commerce
- What we observe in agentic shopping behavior through Inception Trace, our session-correlation layer
- Tactical merchant playbooks for the new surface
- The strategic shifts we think are real, and the ones we think are overhyped
The posture is honest. We will tell you what’s working, what’s not, and where we got predictions wrong. The agentic commerce category is too young for confident pronouncements. The merchants who win this will be the ones who pay attention, run experiments, and update fast.
If you run a store and the last 14 months of ChatGPT Search, AI Overviews, and ACP feel like background noise, they aren’t. They’re the front edge of the most significant shift in commercial discovery since Google.
The buyers are already sending agents. Time to make sure your store knows how to talk to them.