Testing & editorial methodology

MetadataViewer.org is maintained by the site operator. We do not invent individual authors, reviewers, user quotes or credentials. Technical pages identify the behavior we actually test, distinguish detection from cryptographic verification, and link primary specifications or vendor documentation when a claim depends on an external standard.

How file-cleaning behavior is tested

• Byte-level parsers have automated positive, negative and preservation cases.
• Cleaning tests confirm targeted metadata disappears while unrelated structural data remains.
• Every release is built as production HTML and checked for canonical URLs, JSON-LD validity and sitemap consistency.
• High-risk paths are also tested through the browser: inspect a file, download the clean copy, then inspect the copy again.

C2PA tests include multi-part JPEG APP11/JUMBF data, unrelated APP11 segments and PNG caBX chunks. Release acceptance also uses image fixtures from the official contentauth/c2pa-rs project.

Google Takeout date repair tests

Google Takeout photo date repair is tested with little- and big-endian EXIF, timezone offsets, missing capture dates, duplicate and truncated filenames, and corrupt input. Exported copies are independently read back and decoded in Chrome. Tests compare JPEG image bytes and decoded pixels, and check that unrelated metadata stays intact. This is a small-batch JPG/JPEG workflow, not a full archive migration tool.

What a test result does not prove

Metadata is editable. A timestamp is not proof of authorship; a missing C2PA container is not proof that an image is camera-original; and container detection is not certificate or signature validation. Each tool page documents its format-specific limits and fallbacks.

How website meta tags are checked

The meta tag checker makes a bounded HTTP request to a public URL and parses the returned <head>. It follows at most five redirects, rejects local and reserved network targets, stops after 1 MiB or ten seconds, and does not execute the target's JavaScript. Automated tests cover redirect handling, private-address rejection, duplicate/conflicting tags and the distinction between server HTML and tag-like text inside scripts.

Editorial source policy

Standards claims should cite the standards body or reference implementation. Platform instructions should cite the platform vendor. Open-source behavior should cite the project documentation. Time-sensitive statements are dated or removed when they cannot be kept current. Search performance is evaluated with Search Console data, while unavailable data is reported as unavailable rather than treated as zero.

Privacy and corrections

Local files are processed in the browser and are not uploaded to MetadataViewer.org. URL tools are documented exceptions: the YouTube viewer contacts a public metadata endpoint, while the meta tag checker fetches a public web page through the site's Worker. To report a reproducible bug, an incorrect claim or a source that needs updating, email contact@metadataviewer.org with the page URL and a non-sensitive test case.