Discover
Find listings that satisfy category, location, price, availability and fulfillment constraints.
Marketplaces
Keep marketplace choices trustworthy.
Coming later: verify seller, offer, fee and fulfillment facts. Agents stop before cart, enquiry or booking.
For retail, service, B2B, rental and peer-to-peer marketplace teams.
Test the choice between sellers, not only the item.
Marketplace scenarios preserve offer ownership, conditions and trust evidence from discovery through safe review.
Identify
Keep the listing, seller, variant and offer attached to the same source throughout the journey.
Compare
Compare total cost, seller terms, reputation and fulfillment without blending competing offers.
Verify trust
Distinguish platform guarantees, seller claims, reviews and moderation signals.
Stop before commitment
Keep the selected seller and offer up to the cart, enquiry or booking step, then stop.
Recover
Handle delisted offers, changed fees and unavailable sellers without silent substitution.
Keep one offer intact across a multi-party journey.
Independent assertions verify identity, commercial terms and the selected final state.
Qualified listing
The listing satisfies the requested constraints and resolves to a stable seller and offer.
Trusted comparison
Fees, policies, reviews and fulfillment facts remain attributed to the correct party.
Stop point
The intended listing, seller and total are unchanged at the cart, enquiry or booking step, where the run stops.
Marketplace errors can produce a valid item from the wrong party.
Seller confusion
The agent carries one seller's reputation or terms into another seller's offer.
Incomplete total
Platform, service or delivery fees appear only after the comparison decision.
Stale availability
A listing remains discoverable after the selected offer is no longer actionable.
Silent substitution
Recovery changes the seller, condition or fulfillment path without explicit confirmation.
Measure confidence in the selected party and offer.
Completion evidence must show who supplies the result, under which terms and at what total cost.
Step screenshots
Follow discovery, seller comparison and the final check step by step, with masked screenshots.
Completion time
See how long search, filters and each listing step took.
Layout shifts
Measure layout shifts on result, listing and seller pages.
Offer attribution
Assert listing, seller, condition, total and fulfillment at every consequential step.
Public runs stop before payment, enquiry submission or commitment.
AI agents stop before purchases, seller messages, bids, bookings and disputes. Tests of those steps with approved sandbox fixtures are coming later.
The audit inventories marketplace structure and multi-party state.
Established web foundations are evaluated separately from optional agent-native capabilities.
- Product, Offer, AggregateOffer and review structured data
- Seller, listing and canonical identity
- Ranking, filtering and pagination semantics
- Fee, condition, availability and fulfillment state
- Layout shifts during task runs
- Declared WebMCP forms when present: detected, not run
Give marketplace teams evidence tied to each party.
- Results per buyer task for each agent that ran
- Step screenshots and final-state checks
- Seller, listing, fee and policy attribution conflicts
- Recovery findings with reproduction steps and stop gates
- Prioritized backlog with acceptance criteria and retest baseline
Marketplace audit questions
Do public tests contact sellers or place orders?
No. Public runs stop before messages, bids, bookings, payment or another consequential submission.
Can seller and provider portals be tested?
Coming later, with approved test accounts, fixtures and scoped permissions. Agents stop before login today.
Are reviews treated as verified facts?
No. The audit checks attribution and presentation. It does not convert user-generated claims into platform-verified facts.
Practical testing guide
Keep one marketplace offer intact.
Build a repeatable scenario for seller identity, total cost, trust evidence, UGC boundaries, availability recovery and safe transaction gates.
Read the marketplace testing guideRelated
- GuideTest one marketplace offer end to end.Test marketplace websites with AI agents across seller identity, offers, total cost, trust, availability, recovery and safe transaction boundaries.
- MethodologyHow a task is defined, run, checked and attributed.
- Use caseE-commerceAvailable today: AI agent task tests for product discovery, price, stock, delivery and returns, plus the automatic site check. Cart and checkout journeys are coming later.
- Use caseBooking and travelComing later: AI agent task tests for search, availability, prices, policies and timezones. Today every website gets the automatic site check.
- Use caseSaaSComing later: AI agent task tests for SaaS product discovery, documentation and the path to signup. Today every website gets the automatic site check.
- Use caseDigital agenciesRun automatic site checks across client sites with ranked fixes and retests. AI agent task tests for client journeys are coming later.
Test one marketplace choice from query to safe review.
Apply with a listing discovery, seller comparison or provider journey for a private partner review.