UXFixUXFix
Language
Free check · AI search & agents · 12 rules

Can AI shopping agents find and buy from your store?

Paste your URL. A real browser opens your store, adds a product, reaches checkout and judges the 12 agent rules alongside the full 115-rule audit. Free, and nothing is installed.

Free · no card · about 2 minutes · we check it is a real store first · nothing is ordered
What the agent has to manage · demo
/products/…Read the price, the stock and the schema, scripts off
/cartAdd to cart and keep the session, cookies only
/checkoutReach the first checkout step without a CAPTCHA~
/.well-known/…Find an ACP or UCP readiness signal to pay by machine
Demo data, not a measurement. Your report carries your own evidence.
How it works

Find. Read. Buy.

P01

Reads your product page, scripts off

The way AI crawlers fetch it. A price that only JavaScript renders is invisible, and the schema is parsed field by field.

C02

Adds to cart, walks to checkout

A real browser session, then the first checkout step. It records CAPTCHAs, sign-in walls and machine-readable signals on the way.

1203

Judges 12 rules, shows what blocks agents

Pass, partial, fail or not assessable, each with its evidence in the report. Nothing is guessed. The order-confirmation rule always reads as not assessable, because an audit never places a real order.

What we check

The 12 agent rules

From the same rulebook as the full audit. An agent has to find the product, read it without a browser's help, and complete the purchase by machine; the rules follow that chain.

RuleWhat we checkPasses when
PDP-A01Complete Product schema.org: price, availability, GTIN/SKU, rating, shipping detailsproduct pageHighComplete Product schema.org: price, availability, GTIN/SKU, rating, shipping details.✓ pass All required + all recommended fields present.
PDP-A02Price and availability present in DOM as text (not JS/image only)product pageHighPrice and availability present in DOM as text (not JS/image only).✓ pass Price and availability found in server-rendered HTML.
PDP-A03llms.txt present at root; product URLs not bot-blockedproduct pageMediumllms.txt present at root; product URLs not blocked by robots.txt for AI crawlers.✓ pass /llms.txt exists (any content) AND robots.txt does not block AI crawlers from product URLs.
PDP-A04Canonical URLs stable; variant URLs machine-readable and permanentproduct pageMediumCanonical URLs stable; variant URLs machine-readable and permanent.✓ pass Canonical present, absolute, and variants have stable URLs.
CRT-A01Cart state reachable via stable URL or API endpointcartMediumCart state reachable via stable URL or API endpoint.✓ pass Cart URL stable AND JSON endpoint returns cart state.
CRT-A02No session trap that breaks agent-driven cart hydrationcartMediumNo session trap that breaks agent-driven cart hydration.✓ pass Cart URL portable across sessions when logged in; or clearly session-bound with a shareable share-cart link.
CRT-A03Cart line items include machine-readable product IDs/GTINscartLowCart line items include machine-readable product IDs / GTINs.✓ pass Machine-readable IDs present.
CHK-A01Agentic checkout readiness signals (ACP/UCP endpoints declared)checkout step 1HighAgentic checkout readiness signals (ACP/UCP structured endpoints declared).✓ pass ACP or UCP signals present.
CHK-A02No CAPTCHA-only gate without accessible agent pathcheckout step 1MediumNo CAPTCHA-only gate without accessible agent path.✓ pass Either no CAPTCHA OR agent bypass token accepted.
CHK-A03Machine-readable order confirmation (JSON-LD Order schema)checkout step 1MediumMachine-readable order confirmation (JSON-LD Order schema).✓ pass Order schema present with items, total, orderNumber.
CHK-A04Delegated-payment compatibility signals present (tokenization endpoints)checkout step 1MediumDelegated-payment compatibility signals present (ACP tokenization endpoints).✓ pass Delegated-payment signals present.
CHK-A05Structured errors returned in machine-readable format for agent flowscheckout step 1MediumWhen an agent-facing endpoint is declared, errors are returned as structured machine-readable responses (JSON with an error code/field) rather than HTML error pages.✓ pass Error response is JSON with a machine-parsable error code and a field/reason reference.

Best practices

What the stores that pass these rules do, with the source beside each one.

1
Put the schema in the server response, not in a script

AI crawlers fetch HTML and leave. Schema added by a reviews app or a tag manager after load is invisible to them, even though it passes browser-based validators.

EvidenceMost AI crawlers do not execute JavaScript; JSON-LD must be in the initial response. · OpenAI crawler docs · Google Search Central
2
Use the schema.org vocabulary for availability, not words

'In stock' means nothing to a parser. https://schema.org/InStock does. The same goes for conditions and price validity dates.

EvidenceOffers with an invalid availability value are treated as unknown stock and can be dropped from purchasable results. · schema.org Offer · Google Merchant listings
3
Render the price server-side

Currency converters and storefront APIs that fill the price after load leave a blank for crawlers, agents and slow connections. Print the default price in the HTML and let scripts update it.

EvidenceAI shopping agents read the server response; a price rendered only by JavaScript is missed. · OpenAI crawler docs · agent-commerce specs
4
Show stock beside the price, not after a variant pick

'In stock · ships today' next to the price answers the second question a shopper has. Hiding it until a size is chosen delays the answer by a tap.

EvidenceAvailability shown near the price reduces pre-purchase hesitation in usability testing. · Nielsen Norman Group, e-commerce product pages
How common is this? · September 2026
5.1%
of 451 audited stores have a complete Product schema · 74.3% fail outright · 20.6% partial
PDP-A01Complete Product schema.org: price, availability, GTIN/SKU, rating, shipping details
All statistics →
Pass rate per rule · audited stores% pass cleanly
PDP-A02Price and availability present in DOM as text (not JS/image only)85.6%
n=395 · 4.3% fail outright · 10.1% partial
CHK-001Guest checkout offered as first-class option (no account wall)54%
n=226 · 19.5% fail outright · 26.5% partial
PDP-004Shipping cost or estimate discoverable from PDP34.5%
n=383 · 43.6% fail outright · 21.9% partial

Three stores that get it right

All 3 passing stores ›
TThebrick
All required + all recommended fields present.
This store's audit stored no picture for this rule.Audited 2026-09-11
Thebrickshopify61

passes all 2 rules · PDP-A01, PDP-A02

TToms
All required + all recommended fields present.
This store's audit stored no picture for this rule.Audited 2026-09-08
Tomsshopify52

passes all 2 rules · PDP-A01, PDP-A02

MMejuri
All required + all recommended fields present.
This store's audit stored no picture for this rule.Audited 2026-09-08
Mejurishopify65

passes all 2 rules · PDP-A01, PDP-A02

Related articles

Resources ›
Audit findings · 17 min readShipping Cost on the Product Page: How PDP-004 Is Scored

PDP-004 is one of the most-failed critical rules in our audits: here is exactly what a passing product page shows, where it shows it, and how to make the data readable by AI shopping agents.

Read →
Audit findings · 13 min readAll 115 Ecommerce Checkout UX Rules We Audit Against

A full breakdown of the 115-rule checklist UXFix uses to audit product pages, carts and checkout for both human shoppers and AI shopping agents.

Read →
Agent UX · 8 min readWhy ChatGPT and Gemini Don't Recommend Your Store

If your product pages are indexed and correctly priced but AI shopping agents still skip you, the problem is almost always missing schema, an incomplete feed, or no structured reviews.

Read →

Judged in the full audit

The agent rules run inside the free 115-rule audit, on the same walk as the human rules, so one run answers both questions.

Not assessable is an answer

A CAPTCHA or a bot wall stops an agent the same way it stops our browser. When that happens the report says so instead of guessing.

~

Dated, because this moves

What ChatGPT, Gemini and Perplexity can each do changes month to month. Your report is dated, and a re-run shows what changed.

Which AI assistants does this cover?

The rules test the mechanics every agent needs: fetchable pages, complete structured data, a cart and checkout a machine can traverse, and the readiness signals the agent-commerce protocols look for. They are not tied to one assistant's feature list.

Will this place a test order?

No. The walk stops at the checkout form and never submits payment. Nothing is ordered and nothing is written to your store.

What are ACP and UCP?

Two emerging agent-commerce protocols: ways for a store to tell shopping agents how to complete a purchase by machine. The audit records whether your store declares support for either.

Free · no card · about 2 minutes

The 12 agent rules run inside the free audit.

One walk answers both questions: what human shoppers hit, and what stops AI agents. Every failing rule comes with its evidence.