How to read automated and manual statuses
“Pass” applies only to the narrow check described. “Needs review” identifies a signal or recommendation to inspect. Automated and manual checks are kept apart. An automated “Inconclusive” means the check ran but evidence was unavailable or ambiguous. A manual “To verify” means the check cannot be decided from the page: follow its steps and record the result. This tool does not establish AI visibility, real crawler access, rich-result eligibility, policy compliance or working checkout.
Product data compared across sources
Compare exact values across the selected offer, anonymous HTML, visible page and an imported feed. On small screens, scroll the evidence table horizontally. Missing data stays unassessed. A match is not proof of platform acceptance.
Enter the values shown on the page for this variant
Choose a price candidate to inspect, then enter only values you actually verified on the page. Leave unavailable fields blank. Prices use a decimal point; never infer a currency from $ alone.
Import a merchant feed and compare its row
CSV (comma or semicolon), TSV, flat JSON / JSONL, RSS/Atom XML. Up to 5 MiB / 10,000 rows. Import stays local; only the selected normalized row and summary are saved. Export a relevant market/product subset. Platform feed ingestion is not tested.
Feed item IDs, seller offer IDs and SKUs are different identifiers. Without the ID/SKU option, matching uses explicit SKU, GTIN or exact URL. Profiles inspect the selected mapped row, not upload acceptance.
Match feed columns to product fields
Compare this scan with an earlier scan of the same URL
Scan the page again using the extension after a fix. This compares saved snapshots of the exact same URL. Both are re-evaluated using this version's rules. Manual status overrides do not count as fixes. Assisted and imported comparisons remain conditional on the evidence you provided.
Find failures that repeat across your saved scans
How the scan works and what it cannot see
Automated vs manual. 34 checks are automated: the extension decides their status from observed evidence. 28 need manual verification: 3 are calculated once you confirm visible values or import a feed in this report, 18 require a platform account, server logs or a sandbox test, and 7 apply only to optional checkout integrations. Manual checks stay “To verify” until you record a result with evidence, and are never counted as automated results.
Scans read the current rendered page and make anonymous GET requests from its isolated context to the same origin for HTML, robots.txt, discovery metadata, a generated missing URL and a bounded sitemap sample. Visible price snippets are collected with bounded selectors. Same-origin redirects are followed; the browser blocks cross-origin redirect destinations. No forms are submitted, carts changed or external AI services called.
Optional user-agent comparison (AC-14). Off by default. Five extra requests use one unique diagnostic query URL: a browser baseline and four requested crawler user-agent strings. Cookies are omitted. Header rules match only that exact diagnostic URL and tab. Cleanup is verified after each request and before scans, including when the option is off. Diagnostic redirects are not followed. If the query baseline differs from the original page in inspected title/product fields, comparison is inconclusive. Results describe only inspected signals; they do not prove crawler access or page equivalence. Confirm with DevOps using each provider’s documented verification method. Pass/Fail overrides are disabled for this simulated check.
JSON-LD is inspected with local reference resolution. Microdata and RDFa are detected but not parsed. This is a basic inspector, not a complete JSON-LD processor or schema validator. Advanced contexts, external references, shadow DOM, frames, language coverage and session differences may require other tools.
Limits: 2 MiB per response/combined rendered JSON-LD, 512,000 bytes for robots.txt, 64,000 bytes for the missing-URL probe, 12 seconds per request, two initial sitemap files plus up to two child files. Oversized inputs are inconclusive, not failed. Only the newest 20 scans are retained; older scans can be removed earlier to stay within local storage limits.
Each scan has its own URL, timestamp, evidence and manual notes. A new scan does not inherit earlier findings. References are provided for checking current requirements; the ruleset date is 4 October 2026. Program eligibility and availability must be verified at use time.
Reports stay in this Chrome profile. Exports and copied prompts can include page URLs, product data and your notes; review before sharing. Privacy and permissions