AI Commerce Agent Readiness Auditor

Pricing

Rating

Developer

Actor stats

Categories

Share

Audit an e-commerce site for the public signals that make it easier or harder for AI agents, agentic browsers, and LLM-assisted shoppers to discover products, understand availability, find policies, and move toward checkout.

The actor loads each storefront in a real browser, checks machine-readable access, structured product data, search, product/category paths, cart affordances, policy discoverability, and task friction. It returns scores such as agentReadinessScore, machineAccessScore, structuredDataScore, productDiscoverabilityScore, checkoutReadinessScore, policyDiscoverabilityScore, performanceScore, topIssues, and recommendedFixes.

This is an observable readiness auditor. It does not measure proprietary ranking inside ChatGPT, Gemini, Perplexity, Visa, New Gen, Kepler, or any private agent network.

Why use this actor

AI-driven shopping introduces a practical problem for e-commerce teams: a store can look fine to humans while still being awkward for software agents to parse, search, compare, or act on. Product pages may hide variants behind JavaScript-only UI, schema markup may be incomplete, policies may be hard to find, or search may not accept natural-language tasks cleanly.

Use this actor when you need to answer:

• Can a browser-based agent reach the storefront?
• Are products, offers, prices, and availability visible in text or structured data?
• Can a natural-language shopping task find a plausible product path?
• Are shipping, returns, contact, support, and FAQ pages easy to discover?
• Which fixes would improve readiness before a redesign or campaign?
• Which technical fixes should the team prioritize first?

What it produces

For each store and each shopping task, the dataset includes:

• agentReadinessScore: overall 0-100 readiness score.
• readinessLevel: excellent, good, average, poor, or blocked.
• taskOutcome: completed, partial, blocked, or error.
• failureReason: the main observed reason a task did not complete cleanly.
• Score components for machine access, structured data, product discovery, checkout readiness, policy discovery, and performance.
• visibleProductSignals: product, search, cart, price, and schema signals.
• machineAccessSignals: robots, sitemap, llms.txt, canonical, Open Graph, and metadata signals.
• topIssues: the highest-impact issues found.
• recommendedFixes: practical remediation steps.
• evidence: compact counts and samples supporting the audit result.
• screenshotKey: key-value store screenshot of the loaded storefront.

The actor also writes RUN_SUMMARY with per-site summaries and the run average.

Use cases

E-commerce operators can run this before a launch, migration, template change, or merchandising push.

Commerce agencies can audit prospective clients and show concrete technical work.

AEO and SEO teams can add structured data, sitemap, policy, and product-discoverability checks to recurring audits.

Growth teams can monitor whether product discovery and checkout readiness regress after site changes.

Input

The input is intentionally short:

{
"startUrls":[
{"url":"https://demo.saleor.io/"}
],
"shoppingTasks":[
"Find a product suitable as a gift",
"Find shipping or delivery information"
],
"proxyConfiguration":{
"useApifyProxy":false
},
"datasetId":"optional-existing-dataset-id",
"runId":"optional-client-or-workflow-id"
}

Input fields

Output example

{
"inputUrl":"https://demo.saleor.io/",
"finalUrl":"https://demo.saleor.io/default-channel",
"domain":"demo.saleor.io",
"auditStatus":"success",
"shoppingTask":"Find a product suitable as a gift",
"agentReadinessScore":57,
"readinessLevel":"average",
"taskOutcome":"partial",
"failureReason":"Product path was visible but price or cart signals were incomplete.",
"stepsTaken":2,
"durationSeconds":5.1,
"interactionCount":2,
"machineAccessScore":55,
"structuredDataScore":20,
"productDiscoverabilityScore":90,
"checkoutReadinessScore":75,
"policyDiscoverabilityScore":20,
"performanceScore":80,
"visibleProductSignals":[
"product/category links visible",
"storefront search visible"
],
"machineAccessSignals":[
"robots.txt reachable",
"sitemap.xml reachable",
"canonical",
"open graph"
],
"topIssues":[
"Price, availability, cart, or checkout signals are weak before purchase."
],
"recommendedFixes":[
"Make price, availability, variants, and add-to-cart actions visible in text and structured data."
],
"evidence":{
"homepageTitle":"ACME Storefront, powered by Saleor & Next.js | Saleor Store",
"linkCount":37,
"schemaTypes":[],
"taskEvidence":["used storefront search","price text visible"]
},
"screenshotKey":"SCREENSHOT-demo-saleor-io-1780000000000",
"auditedAt":"2026-06-17T12:00:00.000Z",
"runId":"client-q3-audit"
}

API usage

Run the actor from the Apify API:

curl-X POST "https://api.apify.com/v2/acts/YOUR_USERNAME~ai-commerce-agent-readiness-auditor/runs"\
-H"Content-Type: application/json"\
-H"Authorization: Bearer $APIFY_TOKEN"\
-d'{
 "startUrls": [{ "url": "https://demo.saleor.io/" }],
 "shoppingTasks": [
 "Find a product suitable as a gift",
 "Find shipping or delivery information"
 ],
 "proxyConfiguration": { "useApifyProxy": false }
 }'

Fetch dataset rows after the run:

curl"https://api.apify.com/v2/datasets/DATASET_ID/items?clean=true"\
-H"Authorization: Bearer $APIFY_TOKEN"

How scoring works

The score uses observable browser and HTTP signals:

• Machine access: robots.txt, sitemap.xml, llms.txt, canonical tags, Open Graph, descriptions, and blocking.
• Structured data: JSON-LD/schema.org Product, Offer, Organization, WebSite, and breadcrumb signals.
• Product discoverability: product/category links, storefront search, visible product language, and homepage product paths.
• Checkout readiness: price, availability, cart, checkout, and add-to-cart signals visible before purchase.
• Policy discoverability: shipping, delivery, returns, refunds, support, FAQ, and contact links or text.
• Performance: lightweight browser-observed load friction.
• Task simulation: whether the actor can follow the supplied natural-language shopping task to a useful product or policy path.

The scoring model is heuristic and designed for triage. Treat it as a practical readiness audit, not a legal, accessibility, SEO, or paid-agent certification.

Limitations

• The actor checks public observable signals only.
• It does not complete purchases, submit payment details, log into accounts, or bypass access controls.
• It does not call private New Gen, Kepler, Visa, ChatGPT, Gemini, or search-ranking APIs.
• Heavily protected storefronts may require proxy settings or may still return blocked rows.
• Some stores expose product data only after complex client-side interactions, which can lower scores even when a human can browse normally.
• The actor does not guarantee that an LLM will cite, recommend, rank, or purchase from a store.
• Results may vary by geography, site experiments, bot protection, personalization, and catalog changes.

FAQ

Is this the same as New Gen or Kepler Agent Score?

No. It is inspired by the same market shift toward agentic commerce, but it is a TroveVault readiness auditor based on public browser and machine-readable signals. It does not use New Gen's private API or proprietary scoring.

Can I schedule this?

Yes. Use Apify schedules and optionally provide datasetId to append recurring results into one long-running dataset.

Does it support MCP workflows?

The actor produces structured dataset rows that can be consumed by MCP-enabled tools, dashboards, or downstream automations. It does not currently write directly into an MCP connector.

Is scraping these sites legal?

The actor visits public pages and reports observable technical signals. You are responsible for ensuring your usage complies with applicable law, website terms, and internal policies. The actor does not bypass logins, paywalls, or access controls.

Related actors

This pairs well with TroveVault product scrapers, SEO metadata scrapers, website compliance auditors, and competitive monitoring actors when you need a broader e-commerce readiness workflow.

Changes

0.1.0

• Initial.
• Browser-based storefront audit.
• Machine access, structured data, product discovery, checkout readiness, policy discovery, performance, and task-friction scoring.
• One row per store and shopping task.
• RUN_SUMMARY, screenshot evidence, datasetId, and runId support.

Support

Open an issue with a sample storefront URL, the shopping task you expected to work, and the dataset row that looked wrong.