How do AI agents actually shop?
An AI agent does not shop like a human. It does not open a website, does not scroll through categories, and is not swayed by a banner. Instead, it accesses product data directly through structured interfaces and open protocols, compares it, and completes the purchase.
Five protocols shape this process:
ACP (Agentic Commerce Protocol)
An open standard that lets agents handle purchase processes directly, connected to checkout and payment providers.
UCP (Universal Commerce Protocol)
A standard that structures product data so different agents can read it consistently.
MCP (Model Context Protocol)
The technical foundation through which AI models access external data sources and tools in a controlled way.
AP2 (Agent Payments Protocol)
An open standard that lets agents trigger payments securely and with authorization on behalf of the consumer.
A2A (Agent-to-Agent Protocol)
A standard for direct communication between agents, through which an assistant agent and a merchant agent exchange requests, offers, and confirmations.
For you as a shop operator, this means: it is not the prettiest storefront that wins, but the product catalog an agent can reliably read, understand, and classify as trustworthy.
Why does this matter now?
The infrastructure for agentic commerce is not emerging in a few years – it is emerging now. Major commerce platforms and payment providers have begun opening their online shops and product catalogs to AI agents.
Here is the problem: if you are not agent-ready today, you do not lose quality – you lose visibility. An agent that finds three comparable products but can only reliably read and verify two of them will simply not suggest the third. For the consumer, in that moment, it does not exist.
And unlike traditional search engine optimization, there is no second page here where you can still be found. The agent makes a selection – often just a single recommendation. If you are not in that selection, you do not exist.
That is the real urgency: this is not about catching a trend early. It is about not becoming invisible while buying behavior shifts.
How do I know if my online shop is agent-ready?
A quick self-check. Every question you hesitate on marks a starting point – not a verdict.
- Is your product data stored in a structured way (Schema.org, clean, complete attributes), or does it exist only as body text and images?
- Do your product images have verifiable provenance (for example C2PA), or could an agent classify them as potentially misleading or unverifiable?
- Do your images meet current requirements, such as the EU AI Act on labeling AI-generated content?
- Is your shop technically accessible to agents and crawlers, or do robots.txt rules or a purely JavaScript-based rendering block access unintentionally?
The more often you hesitate, the bigger the blind spot – and the bigger the opportunity in fixing it.
What are examples of AI-to-AI commerce?
Imagine a consumer asks their personal AI assistant: “Find me a gift for under 100 francs, deliverable by Friday.” The assistant queries several merchant agents at once for price, availability, and delivery time, compares them, and completes the purchase with the best-fitting provider – without the consumer ever visiting an online shop.
On the other side is your shop agent, answering that request: it returns structured product data, verified images, and reliable availability. If that dialogue between the machines does not happen – because your data is incomplete or your images are not verifiable – your product drops out of the selection.
Visibility is no longer decided only at the moment a human clicks “Buy,” but already in the conversation between two agents that the human never gets to see.
The three prerequisites for agentic commerce
For your online shop to be visible and selectable to agents, three things are needed:
- 1
Structured product data
Machine-readable, complete, standardized (Schema.org, clean attributes).
- 2
Compliant image assets
With provenance (C2PA), EU AI Act compliant, without misleading representation.
- 3
Technical accessibility
Open interfaces, crawler access, connection to the relevant protocols.
The first and third prerequisites are familiar to many from traditional technical optimization. The second – compliant, verifiable image assets – is new and most often overlooked. This is exactly where Product Asset Studio comes in.
Traditional online shop vs. agent-ready online shop
| Attribute | Traditional | Agent-ready |
|---|---|---|
| Product data | Body text, inconsistent | Structured, Schema.org |
| Image provenance | Not verifiable | C2PA manifest, EU AI Act compliant |
| Discoverability | Optimized for humans | Optimized for humans and agents |
| Visibility to agents | Random to nonexistent | Predictable |
The solution: Product Asset Studio
Of the three prerequisites, one is especially demanding – and most often underestimated: compliant, verifiable image assets at the volume a real product catalog requires.
Product Asset Studio handles exactly this part. From a single product image, it creates a complete, verified asset set – each image rated with a Fidelity Score and equipped with a C2PA provenance record. This makes your product images not only appealing to humans, but traceable and trustworthy to agents.
That is the image-asset side of agent-readiness – automated, compliant, and scalable.
Conclusion
Agentic commerce is not a future scenario but an infrastructure already taking shape. Online shops that are not agent-ready today do not lose quality – but they lose visibility, exactly where buying decisions are increasingly made. The first, often overlooked step begins with your product images.
TL;DR
- Agentic commerce means shopping through autonomous AI agents (ChatGPT, Gemini, Perplexity), without individual human steps.
- Agents often choose only one recommendation – if you are not agent-ready, you do not show up.
- An agent-ready online shop needs three things: structured data, compliant image assets, technical accessibility.
- Product Asset Studio delivers the image-asset side: verified assets with a Fidelity Score and C2PA record.
Remo Girard
Co-Founder of LumAIa
Remo Girard is Co-Founder of LumAIa and CEO of iundf Dewave, a Swiss digital and AI transformation agency for SMEs. He has worked for years on e-commerce, custom AI solutions, GEO/AEO, and the agentic web, and guides Swiss companies through the shift to AI-powered marketing and commerce.