Agentic Commerce, Defined: The 7-Layer Map of How AI Buys

Most vendor pitches in this category sound interchangeable because they're describing the same problem from one layer of a seven-layer stack. The map sorts the noise. After you read it you can tell who owns what — and where most brands silently lose money.

Where this category actually lives

Agentic commerce is the discipline of getting an AI agent — ChatGPT, Gemini, Perplexity, Rufus, Copilot, Claude — to discover, recommend, and complete a purchase on a shopper's behalf. The shorthand is everywhere now. The infrastructure under it is not.

What's missing in the public conversation is a clean map of the stack. The pitches sound interchangeable because every vendor talks about the same end-state — "an AI agent buys your product on a shopper's behalf" — without telling you which layer they own. That's the gap this post closes.

There are seven layers. Each does a different job. Each has a different set of vendors. The unspoken rule in this market is that the layer you own is the layer you defend. We think that rule is wrong — and the rest of this post is why.

The 7-Layer Map

Read it the way you read a network diagram. The AI surface at the top fields the shopper question. The protocols at Layer 02 carry the answer back and forth. Payments and card issuance at 03/04 turn an agent's "yes" into a settled charge. Checkout at 05 actually finishes the order. Discovery at 06 is what got the brand into the answer in the first place. Trust at 07 is the layer most vendors don't talk about until something breaks.

The category looks crowded because there are 30+ named vendors on the map. It is in fact uncrowded — per layer — because almost every vendor sits in exactly one cell.

A simple shopper journey, layer-by-layer

The map gets practical when you trace one transaction through it.

Take the case of a shopper who asks ChatGPT, "What's the best non-toxic frying pan under $200?"

1. Layer 01 — The shopper's question arrives at OpenAI's surface. ChatGPT decides this is a transactional query and routes it to its shopping subsystem.
2. Layer 06 — ChatGPT pulls candidate products from indexed sources. This is where your brand either appears or doesn't. If you've optimized for AI discovery, your SKU is in the candidate pool. If not, you are invisible.
3. Layer 02 — ChatGPT formats the recommendation using the Agentic Commerce Protocol (ACP). The product card it shows the shopper carries a structured payload describing your SKU, price, availability, shipping, return policy.
4. Layer 05 — The shopper taps "Buy." ChatGPT sends a checkout intent through ACP. Your Tru Commerce integration receives it, validates the cart, locks the inventory, and returns a checkout confirmation — all without the shopper leaving the conversation.
5. Layer 03 + 04 — A tokenized agent-payment instrument settles the charge. Your payment processor sees a normal-looking transaction; the agent layer is invisible to the merchant of record.
6. Layer 07 — Behind the scenes, the surface verifies the agent's identity, validates the payment token isn't reused, and runs fraud signals. If anything fails, the transaction never reaches you.

That is one transaction. Every single layer had to do its job for the order to land. Most brands have two or three of those layers wired. The other four are silently failing.

Where each competitor actually sits

Naming names matters, because the category is full of overlapping pitches that are not in fact overlapping at the layer level.

Layer 06 only (Discovery / merchant enablement):

• Nudge — measurement-led; AEO dashboards
• FERMÀT — shoppable on-site funnels; on-domain conversion
• Wildcard — SKU-level AEO + open agents.json spec
• Catalog — product feed structuring
• Swap Commerce — apparel/returns-first agentic surface
• Profound, Rankly — visibility monitoring

Layer 05 only (Checkout execution):

• Rye — universal product API + checkout (US-focused, dev-led)
• Zinc, Henry Labs — Amazon-adjacent agent checkout
• CartAI — publisher/affiliate-side agentic checkout
• Induced AI — general browser-driven completion

Layer 03/04 only (Payments / issuance):

• Nekuda, Skyfire — delegated-spend tokenization
• Basis Theory — tokenization infrastructure
• Prava, PayOS — compliant agentic payments
• Lithic, Marqeta — virtual card rails

Layer 02 work but no merchant-facing product:

• Stripe, Visa, Mastercard, PayPal, Adyen — protocol contributors

Tru Commerce position: We bridge Layer 02 + Layer 05 + Layer 06 in a single platform. One integration. One contract. The brand never directly manages the six protocols (ACP, UCP, AP2, MCP, A2A, TAP) — we translate. The brand never picks a checkout API per surface — we route. The brand's Citation Rank, Share of Voice, and Visibility Score (the Layer-06 measurement work) feed the same data spine that closes the loop at Layer 05.

That is the entire product thesis in one paragraph. Everything else is a feature.

Why most brands try to build this themselves and regret it

The instinct, when a brand reads the 7-Layer Map for the first time, is to assemble it themselves. Pick a Layer-06 visibility tool. Pick a Layer-05 checkout API. Pick a Layer-03 payments tokenizer. Glue with internal engineering.

We have watched this play out. Three things go wrong.

The protocols change weekly. When ACP added multi-item carts in Q4 2025, every brand that had wired against the single-item v1 schema was suddenly behind a glass wall. UCP shipped catalog endpoint + loyalty endpoints in Q1 2026. Brands keeping up with six protocols on six release cadences burn 1–2 engineers permanently on translation work that doesn't differentiate them.

Attribution falls between the cracks. Layer 05 knows the conversion happened. Layer 06 knows the agent surface that drove it. If those two are different vendors with different IDs and different export formats, the brand never knows which agent surface produced which dollar. The MyMuse case is the canonical example: GA4 reported ₹11.5K/month from AI surfaces. The real number, once the attribution was rebuilt with proper Layer-02/05/06 stitching, was ₹81.2K. A 7× gap, hidden because the layers weren't talking. We wrote this up at length in Dark Agentic Commerce Traffic (DACT).

The "merchant of record" question gets messy. Multiple vendors in this stack want to be the merchant of record for the agent purchase — because it's the position with the most pricing power. A brand that hands MoR to a checkout vendor at Layer 05 quietly loses customer data, pricing control, and the ability to issue a refund in their own brand. That is a strategic loss disguised as a tactical integration.

A single layer wired well beats four layers wired badly. A single integrated platform across the three load-bearing layers beats either.

The two layers that decide the outcome

If you only have budget for two layers this year, pick Layer 06 and Layer 05. Here's why.

Layer 06 is the demand-side gate. If you are not in the candidate pool the agent considers, no amount of checkout wiring saves you. Visibility comes first because nothing else fires until visibility fires.

Layer 05 is the supply-side gate. If the agent decides to recommend you and the checkout drops the shopper into a desktop form, your conversion rate falls by roughly an order of magnitude. (Shopify's recent webinar series quoted 49% higher conversion for AI shoppers when checkout completes inside the agent surface. Our internal measurements on a small cohort of brands sit in the 35–55% range depending on category.)

The remaining layers — protocols (02), payments (03), card issuance (04), trust (07) — matter, but they're table-stakes in the sense that almost every Layer-05 or Layer-02 vendor either provides them or integrates a partner that does. You can usually defer thinking about them.

The two-layer focus is also the budget-realistic move. A brand with $50K–$200K of annual platform spend in this category can cover Layer 06 + Layer 05 well. The same budget split across 04 vendors covering all seven layers buys you four lawn-mowers that don't talk to each other.

How to use the map this quarter

Three actions, in order, for a brand reading this for the first time.

1. Find yourself on the map. What's your current Layer-06 setup? (A monitoring tool? A feed optimizer? Nothing?) What's your Layer-05 setup? (Native to your Shopify checkout? A custom agent checkout API? Nothing?) The honest answer is usually "Layer 06 partially, Layer 05 not at all."

2. Run a free Citation Rank scan. Drop your URL. In 24 hours you'll see how you appear (or don't) across the six AI surfaces — that's your Layer-06 baseline. Three prescriptive recommendations come back with it. Free. No credit card.

3. If the citation rank surfaces a real gap, talk to us. Book a demo — we'll walk through what your stack would look like with Layer 02 + 05 + 06 unified, what changes for your conversion rate, and what the contract looks like. Pricing is public: free tier, then 2% per agent transaction above 10/month. Enterprise pricing kicks in at volume.

The 7-Layer Map isn't the whole story of agentic commerce. It's the part that tells you where to put your money.