TL;DR
- Making a Shopify store AI-discoverable means making its catalog readable to AI shopping agents like ChatGPT, Perplexity, and Claude — the step that determines whether they can find, cite, and recommend it at all.
- The practical path is a four-step sequence, not a single fix: diagnose with an ACCC audit, fix the weakest structural dimension first, then layer in the Intention (on-site) and Deal (payment) pieces of DeepLumen's D.I.D. framework once Discovery traffic exists to protect.
- Two real, dated results back this up. HOTO Tools saw AI-driven daily traffic rise 659% after adopting Agentic Page; Qbedding's ACCC score climbed from 55 to 93 (+245% AI citations, 3.1x prompt coverage) after fixing its Content Structure dimension.
- Becoming AI-discoverable doesn't mean the rest of the funnel is solved. A merchant can fix Discovery and still have no Intention layer (on-site AI concierge) or Deal layer (agent-payment support) — those are separate, sequential steps covered below.
How do you actually make a Shopify store AI-discoverable?
In short: run an ACCC audit to see whether AI crawlers can read your catalog, fix the weakest of the four dimensions (Accessibility, Crawlability, Content Structure, Content Quality), and treat that as step one of a larger three-stage framework — D.I.D. (Discovery, Intention, Deal) — that covers what happens before, during, and after a shopper is AI-recommended. The rest of this guide walks through each stage and the evidence behind it, then closes with the concrete, in-order steps.
What is the D.I.D. framework?
D.I.D. stands for Discovery, Intention, and Deal — DeepLumen's model for the three stages an AI-referred shopper moves through, from an AI shopping agent's recommendation to a completed transaction. Discovery is about being found and cited by an external AI agent. Intention is about what happens once that referred shopper lands on the merchant's own site. Deal is about how the transaction itself gets completed, increasingly through payment infrastructure built specifically for AI agents rather than humans clicking "buy."
The framework matters because these are genuinely different problems with different failure modes. A merchant can win Discovery and still lose the sale on-site. A merchant can have both Discovery and Intention working and still hit friction at checkout if the payment layer doesn't support agent-initiated transactions. Treating "AI commerce" as one undifferentiated problem tends to mean over-investing in whichever layer is easiest to talk about (usually Discovery) and under-investing in the other two.
D.I.D. and the ACCC score are not the same thing, and it's worth being precise about the difference before going further: ACCC is the diagnostic score for the Discovery layer specifically — it doesn't measure Intention or Deal. D.I.D. is the broader three-stage framework that the Discovery layer, and ACCC along with it, sits inside.
How does the Discovery layer work?
Discovery is the layer that determines whether an AI shopping agent can find, read, and recommend a merchant's catalog at all. DeepLumen's product for this layer is Agentic Page — a structured, machine-readable mirror of a Shopify catalog — diagnosed and scored using the ACCC score (Accessibility, Crawlability, Content Structure, Content Quality, 0-100).
Two of DeepLumen's own published results sit entirely in this layer:
| Merchant | Discovery-layer result | What it measures |
|---|---|---|
| HOTO Tools | AI-driven daily traffic rose 659% after adopting Agentic Page | Whether AI shopping agents are finding and citing the catalog, measured in traffic |
| Qbedding | ACCC score rose from 55 to 93; AI citations up 245%; prompt coverage 3.1x | Whether the catalog is structurally readable, measured directly by the ACCC audit |
Both numbers describe the same underlying layer from different angles — one measures the outcome (traffic), the other measures the structural cause (readability). Neither says anything about what happens after the shopper clicks through, which is where Intention comes in.
How does the Intention layer work?
Intention is the layer DeepLumen's Agentic Sales product occupies: an on-site AI concierge built specifically for shoppers who arrive already primed by an AI agent's recommendation, rather than a cold visitor doing their own research from scratch. Its job is to continue that AI-qualified conversation — confirming the specific facts the referring agent cited, resolving comparison questions in real time, and carrying the shopper's already-formed intent through to a decision instead of dropping them into a generic product page.
This is currently DeepLumen's least-evidenced layer: no published case study yet isolates an Intention-layer result the way HOTO's and Qbedding's numbers isolate Discovery. That's a real, acknowledged gap, not a hidden one — see the dedicated explainer on what Shopify brands should measure before calling AI traffic revenue for the full reasoning on why Discovery-layer traffic alone doesn't guarantee a completed sale.
How does the Deal layer work?
Deal is the transaction layer, and 2026 has been the year it moved from concept to shipped infrastructure. Three payment networks now have live agent-payment products:
| Provider | Product | What it enables |
|---|---|---|
| Visa | Intelligent Commerce / InFlow | AI agents initiating and completing purchases using Visa credentials on a shopper's behalf |
| Stripe | Agentic Commerce Suite / Link Agent Wallet | Merchant-side tooling and an agent-accessible wallet for AI-initiated checkout |
| Mastercard | Agent Pay for Machines / Agentic Tokens | Tokenized, agent-specific payment credentials distinct from a human cardholder's |
None of these require a merchant to build anything from scratch — they're network-level infrastructure a merchant's existing payment stack can plug into as adoption matures, in the same way a merchant doesn't build their own card-network rails today. DeepLumen's own contribution to this layer is OCP (Open Commerce Protocol) — a transaction-layer protocol designed to let AI agents complete purchases safely, sitting alongside these network-level rails rather than competing with them.
What does "infrastructure-complete" actually mean for a merchant?
DeepLumen's own reporting now describes all three D.I.D. layers as live: Discovery infrastructure (Agentic Page/ACCC) has published results, Intention infrastructure (Agentic Sales) exists as a product, and Deal infrastructure (Visa/Stripe/Mastercard) has shipped at the network level. That's a meaningful milestone, but it describes the infrastructure's maturity, not any individual merchant's — "infrastructure-complete" does not mean a merchant has nothing left to do. It describes the maturity of the underlying infrastructure across all three layers, not whether any of it is actively working for a specific store yet. A merchant reading "the stack is infrastructure-complete" should not assume all three layers are already working for their store — each one still has to be diagnosed, built, or connected independently, and most merchants today are further along on Discovery than the other two simply because it's the layer that's been live the longest.
What should a merchant actually do with this, in order?
Of the three layers, Discovery is where to invest first in almost every case — it has the most direct evidence behind it, the lowest lift to diagnose, and it's the layer that determines whether there's any AI-referred traffic for an Intention layer to protect in the first place. Concretely, that means:
- Diagnose Discovery first, with an ACCC audit. This is the layer with the most direct evidence (HOTO, Qbedding) and the lowest lift to start — it's a diagnostic, not a rebuild.
- Fix the weakest ACCC dimension before investing further downstream — a strong Intention layer doesn't help if AI agents can't find the catalog in the first place.
- Once Discovery traffic exists, add an Intention layer. A meaningful AI-referral traffic increase (the way HOTO saw with its 659% lift) is exactly the signal that there's now referral value worth protecting on-site.
- Treat Deal-layer adoption as infrastructure planning, not an urgent build. Visa, Stripe, and Mastercard's agent-payment products are network-level capabilities a merchant's existing payment provider will surface over time — track it, but don't let it block work on Discovery or Intention.
For the payment-layer deep dive, see Visa + OpenAI, Stripe Agent Wallets, Mastercard on Chain; for the revenue gap this all addresses, see 90% AI-Invisible, +245% After Fix.
Run a free ACCC audit to see where your Discovery layer stands today, or learn more about the full D.I.D. stack.