Public Product Comparison Evidence Agent

Turn public compare, versus, and side-by-side product pages into structured evidence rows for buyer research, competitive intelligence, and AI-agent workflows.

The Actor focuses on explicit side-by-side evidence, especially public comparison tables. It extracts compared product hints, comparison criteria, table evidence rows, supporting narrative claims, winner hints, pricing mentions, calls to action, risk labels, stable hashes, and change status. It uses lightweight public HTTP fetching only. It does not log in, submit forms, follow private-network URLs, render browsers, or fetch query-token URLs.

When To Use It

• You have a public X vs Y, compare X and Y, or side-by-side product table.
• You want structured evidence instead of a raw web page summary.
• You need to monitor whether a comparison page changed without paying for unchanged repeats.
• You want useful-result pricing: failed, unsafe, generic, focus-mismatched, and unchanged records are written without the useful-result event.

Input

• comparisonPageUrls: public HTTP/HTTPS comparison page URLs.
• focusProducts: optional product or brand terms that must appear before a useful event is charged.
• focusCriteria: optional comparison criteria terms that must appear before a useful event is charged.
• previousComparisonRecords: previous dataset records from this Actor. If the productComparisonHash is unchanged for the same URL, no useful event is charged.
• requestTimeoutSecs: public page request timeout.

Output Highlights

• comparedProductHints
• comparisonCriteriaHints
• comparisonEvidenceRows
• winnerClaimHints
• pricingMentionHints
• biasDisclosureHints
• comparisonEvidenceScore
• productComparisonHash
• changeStatus

Pricing Event

The useful-result event is:

useful-product-comparison-evidence-result

The Actor should not charge merely because an error or partial record is written to the dataset.