Preserving Source Dates, URLs, Metadata, and Document Integrity

A URL is a location, not a preserved source#

Web content changes, moves, redirects, loads dynamically, or disappears. A live link does not preserve what a page said on the date it was used. Capture the complete page in a durable form while retaining the URL, page title, publisher, author when shown, visible publication and update dates, access timestamp, and capture method. If the page relies on tabs, expandable text, embedded documents, or interactive charts, confirm that the saved record includes the content supporting the claim.

Preservation does not make a source reliable; it makes the reviewed version identifiable. A preserved press release remains a press release. A saved ranking still requires methodology. A screenshot without surrounding context may remain weak. Evaluate authority, independence, relevance, and probative value separately. Integrity controls ensure that later reviewers assess the same record rather than a changed webpage or undocumented excerpt.

Create a preservation envelope#

Separate the dates that a source can carry#

A single page may display an event date, publication date, last-updated date, copyright year, and access date. These are not interchangeable. Record each in its own field. If no publication date appears, say “no visible publication date” and retain the access date; do not infer a date from search-result snippets, file paths, or copyright notices without explanation. For a PDF, inspect the document itself and the hosting page. Embedded metadata can provide leads but may reflect software creation or later processing rather than public issuance.

When a page changes, preserve both versions if the difference matters. Record the first capture, later capture, observed change, and which version supports the filed claim. Do not overwrite the earlier file. If an official source corrects an error, acknowledge the correction and update downstream statements. A version log should make clear which record was relied on at each drafting milestone.

Reject filename-based trust#

• ○Open the file and confirm type, completeness, readability, and expected content.
• ○Record the source location and ordinary creator rather than relying on the filename.
• ○Preserve the original extension and MIME-consistent format.
• ○Check page count, attachments, signatures, stamps, tables, and reverse sides.
• ○Use server-generated or controlled working filenames without changing the underlying source.
• ○Calculate integrity values on the untouched capture, not only the final packet.
• ○Store originals read-only and derivatives separately.
• ○Document lawful redactions and verify that hidden text is not recoverable.

Use checksums as change detectors#

A cryptographic checksum such as SHA-256 produces a value tied to the file’s exact bytes. Recalculate it after transfer to detect corruption or unrecorded modification. A checksum does not prove that the source was truthful or that the person who supplied it was authorized; it proves only that the bytes match the preserved reference. Record the algorithm, value, file ID, and capture date in the source register. Keep the value internal unless there is a reason to display it.

Normal processing changes bytes. OCR, rotation, redaction, compression, page numbering, translation, and PDF merging will create different checksums. Preserve the original checksum, assign derivatives explicit version IDs, and log the transformation. Never claim that a redacted or compressed derivative is byte-identical to the source. The chain should let a reviewer move from filed exhibit to derivative to untouched capture without ambiguity.

1. Acquire

   Obtain the complete record from an identified source using a documented method.

2. Capture metadata

   Record title, publisher, author, URL, dates, access time, language, and context.

3. Freeze original

   Store the untouched file, calculate a checksum, and restrict unlogged editing.

4. Create derivatives

   Generate OCR, translation, redaction, or filing copies under new version IDs.

5. Map and cite

   Connect the controlled source and pinpoint to specific factual propositions.

6. Revalidate

   Before filing, test URLs, checksums, page references, and source-version alignment.

Capture web sources reproducibly#

Prefer the publisher’s canonical page or official downloadable document. Remove tracking parameters from the recorded citation while retaining any parameter required to reach the content. Save a PDF with headers and footers that show title, URL, and capture date, but verify those additions do not cover content. When print output omits images, tables, or expanded text, supplement with complete screenshots and describe the relationship. Record redirects and the final resolved URL.

Archive services can help demonstrate an earlier public version, but preserve the archive URL, original URL, archive timestamp, and access date. Understand that archive coverage may be incomplete and that blocked resources or dynamic elements can be missing. Do not represent an archive timestamp as the original publication date. When the live official page remains available, preserve it too and explain why the historical version matters.

Preserve structured and machine-generated evidence with definitions#

Exports should be bounded and reproducible. State whether data reflect all records or a filtered subset. Preserve the query, filters, date range, timezone, and definitions. For screenshots of a dashboard, include the controls that define the metric. For email, preserve enough thread context and header information to identify participants and chronology. For repository data, distinguish authorship of code, review, management, and ownership; a commit count alone may not establish substantive contribution.

Manage broken links and inaccessible sources#

At final review, test every external URL and note redirects, access restrictions, and changed content. If a source is no longer live, rely on the preserved copy and explain its provenance; add a credible archived version when available. Do not replace a broken source with a different page merely because it reaches a similar conclusion. Update the claim map and citations whenever the source changes. The exhibit index and cross-reference guide provides controls for propagating those changes.

Source preservation should begin during evidence development, not at filing assembly. Award rules may disappear after a cycle; staff pages change; metrics update; contracts become inaccessible; and publication pages are redesigned. The evidence inventory guide can assign capture owners and deadlines while material is still available. For translated web or document sources, apply the additional controls in Foreign-Language Evidence and Certified Translations.

Audit the final record for integrity and restraint#

Sample high-consequence claims and trace each backward: filed citation, derivative exhibit, controlled original, checksum, capture metadata, publisher, and proposition. Confirm that the cited passage exists and that the preserved context does not qualify the claim. Review transformations and redactions. Search for stale versions and duplicate source IDs. If a capture cannot be authenticated beyond a weak screenshot, qualify the claim and seek corroboration rather than hiding the limitation.

Integrity is a process claim, not an assertion that every source is true. A well-preserved record may still be biased, incomplete, or promotional; those issues belong in source evaluation. Preservation ensures that the team knows exactly what was reviewed, can detect unintended changes, and can reproduce citations later. That foundation supports better analysis, cleaner updates, and a more credible educational or filing record.

Storage resilience is part of integrity. Keep controlled originals in a location with access controls, backups, and recovery testing rather than on one person’s device. Limit edit permissions, log replacements, and avoid synchronization tools that silently overwrite versions. Retention should be proportionate to lawful business and filing needs, with special care for third-party personal data. A backup is useful only if the team can restore the correct source version and reconnect it to its metadata and claim mappings.