How does an agentic commerce readiness check find where agents drop your shop?
The Buyer Journey Check is our agentic commerce readiness check and audit. We give ChatGPT, Perplexity, Claude and a test agent the shopping tasks your customers would, record where each one stops naming, comparing or buying your products, from the first question to the reorder, and hand your developers the fixes in order of the sales they recover.
What are the five stages evaluated during the audit?
The audit follows the five stages of a purchase made through an assistant: whether it finds your shop, understands the product, can compare your offer, can complete checkout and can come back for the next order. Each stage gets a score, the evidence behind it and a list of fixes ordered by the sales they recover.
-
Discover
- Search and shopping assistants are allowed to read your pages
- Your products show without JavaScript having to run
- Assistants name you when customers ask about your category
-
Understand
- Each product, its offer and your company are described as data (Product, Offer and Organization schema)
- Every product carries its barcode number (GTIN), brand and model, so it matches reviews elsewhere
- The price an agent reads is the price a person sees
-
Choose
- Shipping cost and delivery time are stated as fields an agent can compare
- Your return terms can be read by an agent
- Your ratings can be read by an agent
-
Buy
- An agent can get from cart to paid order
- The checkout connections open in your market are in place
- Text on your product pages cannot be used to hijack an agent (prompt injection)
-
Retain
- An agent can check the status of an order
- An agent can reorder or renew a subscription
- Agent orders show up in your analytics
What did the same checks find on 240 Swiss online shops?
In July 2026 we ran the readiness checks on 240 online shops in French-speaking Switzerland. Most lost agents at the Understand stage, because price, stock and identity were written for people only. The full Romandie audit breaks the results down by stage and by sector.
| What was measured | Source, date and scope |
|---|---|
| None of 240 online shops in French-speaking Switzerland scored an A for agent readiness. The median score was 57 out of 100, and 104 shops scored D or F. | Autobasket audit of 240 Romandie online shops, July 2026. Switzerland, Romandie, 240 shops |
| For 157 of 211 shops (74%), the stage that lost the most agents was reading the product itself: price, stock and identity were not declared as structured data. | Autobasket audit of 240 Romandie online shops, July 2026. Switzerland, Romandie, 211 shops read in full |
| On 117 of 211 shops (55%), the only price an agent could find was in the page text, with no Offer price in the structured data to confirm it. | Autobasket audit of 240 Romandie online shops, July 2026. Switzerland, Romandie, 211 shops read in full |
| About 34% of individual retail product pages cannot be properly accessed by AI tools. | Adobe, AI Content Visibility Checker, reported by TechCrunch, 16 April 2026. Retail product pages tested by Adobe |
What makes an agent drop one bike shop and buy from another?
Asked for a gravel bike under CHF 4,000 for alpine rides, available in Switzerland, an agent checks four things on every shop. It drops shops that load the price with JavaScript, show the geometry as an image or describe delivery vaguely, and it recommends and buys from shops that state those facts as data it can read.
| What the agent checks | Shop that is dropped | Shop that is bought from |
|---|---|---|
| Price under CHF 4,000 | Price loaded by JavaScript, absent from the page itself | Price stated in the page code as price: 3890.00, priceCurrency: "CHF" |
| Suited to long climbs | Geometry chart as an image | Geometry and gearing in an HTML table |
| The same model as reviewed elsewhere | No barcode number (GTIN) | Barcode number, brand and model declared |
| In Switzerland within a week | "Fast delivery" in the footer | Two to three days to Switzerland, declared in shippingDetails |
What is included in the audit package?
You receive four documents: a one-page scorecard per stage for decision-makers, a report on how often assistants name you and who they name instead, an analysis of which agent checkout routes your market offers and which you lack, and a 30-day Technical Roadmap. Every finding comes with the change that fixes it, down to the code.
Readiness scorecard
Where you stand at each stage, the evidence behind it and what to fix first.
AI visibility report
How often each assistant names you when customers ask, and who it names instead, with the margin of error on every rate, measured as described in how to measure share of model.
Protocol gap analysis
Which ways of letting an agent check out your channels need, from MCP and UCP to ACP and AP2, and which ones your shop is missing today, explained in UCP vs ACP vs AP2 vs MCP.
30-day technical roadmap
Tickets your developers can pick up, each with the code change it needs.
How do you commission a Buyer Journey Check?
Send us your shop address and the three categories that matter most to your revenue. We confirm the scope, the missions we will run and a fixed fee, then run the audit on public data, with no access to your systems needed, and walk your team through the findings when the report is ready.
- You send the shop address and your priority categories, which follow your industry or business model.
- We reply with the missions, the assistants tested, the timing and a fixed fee.
- We run the audit from the outside, using public pages only.
- We present the findings to your team and hand over the four documents.
What do merchants ask before booking the audit?
E-commerce and marketing leads usually ask four things before booking: what a buying mission is, what they hold at the end, whether we need access to their systems and how the fee is set. The answers below give you the shape of the audit before we speak.
What is a buying mission?
A task a customer would plausibly give an assistant, such as finding a gravel bike under CHF 4,000 in size M delivered before a given date. We run each mission through ChatGPT, Perplexity, Claude and a test agent connected through MCP, the standard agents use to query a shop directly, and record where it stops.
What do we receive at the end?
A score per stage with the evidence behind it, the missions we ran and where each one stopped, and a list of fixes ordered by the sales each one recovers, written so your developers can work from it directly.
Is my website ready for AI shopping agents?
Three signals give a first answer in minutes: your robots.txt lets search and shopping crawlers in, each product page declares a Product with an Offer carrying the price, currency and availability, and that price matches the one a person sees. In our July 2026 audit of 240 Romandie shops, none passed every stage, and 117 of the 211 we read in full showed their price only as page text.
How is the Buyer Journey Check different from a free readiness score?
A free score, such as the product page check on our home page, reads the code of one page and lists what blocks an agent. The Buyer Journey Check runs your customers' shopping tasks through the assistants themselves, measures how often they name you and your competitors with a margin of error, tests the checkout routes open in your market and ranks the fixes by the sales they recover.
Do you need access to our systems?
No. The audit reads your shop from the outside, the way an agent does, using public pages and the assistants themselves. Access is needed only later, if you go on to the Machine Readability Sprint.
How is the fee set?
As a fixed fee per shop, based on the number of categories and markets we test. You receive it in writing before anything starts.
See your shop the way an agent does
The free check reads one product page in a minute. The Buyer Journey Check covers your categories, your competitors and every stage of the purchase, so your developers fix the problems that cost you the most sales first.