How to preserve and document a web source

A screenshot preserves pixels. It does not, by itself, preserve the URL, retrieval time, redirects, page structure, downloadable file or enough context for another person to reproduce what you saw.

A defensible web-source record is a small evidence bundle: source metadata, contextual captures, an independent archive when appropriate, integrity information for saved files and a log that says exactly what each item proves.

Capture the source before analysing it

Start an evidence log whilst the live page is open. Record:

• the URL you entered and the final URL after redirects
• retrieval date and time with time zone, preferably UTC
• page title, site or publisher, named author and visible dates
• the canonical URL declared by the page, when available
• the claim or passage you are preserving and where it appears
• links, document IDs, video IDs, page numbers or timestamps needed to return to the same place
• whether the page required a login, subscription, interaction or regional access

Copy the relevant text into your notes without silently correcting spelling or punctuation. Put omissions in brackets or mark them explicitly. Keep enough surrounding text to show whether the passage is a quotation, headline, allegation, correction, caption or response to someone else.

Save more than one representation

No single format captures every useful property. For consequential research, consider a proportionate bundle:

1. Context screenshot. Include the address bar or otherwise bind the image to the URL, plus the title, relevant passage and visible date when possible.
2. Full-page capture or print-to-PDF. Useful for reading order and broader context, but inspect it because print styles can hide navigation, comments, interactive charts or expanded text.
3. Original downloadable file. Save a linked PDF, image, dataset or transcript as delivered; do not replace it with a screenshot of the file.
4. HTML or web-archive capture. Useful when links, structure or replay matters, but script-rendered content and remote assets may not save cleanly.
5. Notes about interaction. Record which tab, filter, disclosure or video timestamp exposed the evidence.

Keep the first saved copy unchanged. If you crop, annotate, redact, transcode or extract text, make a working copy and record the transformation. An annotation should never become indistinguishable from the captured original.

Create an independent archive when appropriate

A third-party archive can provide a stable public reference independent of your device.

The Wayback Machine's Save Page Now accepts a page URL and returns a permanent archived URL when the save succeeds. The Internet Archive explains that it saves the single page submitted, not the whole site or the pages linked from it. Perma.cc similarly creates a stable record for a submitted page and advises users to review the finished record to confirm that it displays correctly.

After archiving:

• open the archived URL in a fresh browser context
• check that the relevant text, image, date and layout are present
• test whether an embedded document, audio, video or expanded section survived
• record the archive URL and its capture timestamp
• preserve the live URL alongside the archive URL

An archive failure is not evidence that a page never existed. Login walls, robots rules, scripts, streaming media, rate limits and transient errors can all prevent a complete capture. Perma.cc also documents that it preserves the targeted page rather than recursively preserving every linked page. Archive each essential linked source separately.

Do not use a new archive capture to rewrite the past. It proves what the archive captured at that time, not what the page contained yesterday or when it was first published.

Record file integrity with a hash

For locally saved files, calculate a SHA-256 digest and place it in the evidence log beside the filename, size and capture time. NIST's Secure Hash Standard specifies SHA-256 amongst the algorithms that generate message digests used to detect whether data has changed since the digest was generated.

The interpretation must stay narrow: if the same file later produces the same digest, its bytes are unchanged relative to the recorded copy. A hash does not prove who created the file, whether the live page was authentic, whether your computer clock was correct or whether the content was true.

Generate the digest as soon as practical, keep the original read-only when your workflow allows it and calculate a new digest for every transformed copy. Do not overwrite the original hash after editing a file.

Use an evidence log another person can audit

One row per item is easier to verify than a folder of unexplained screenshots:

Use consistent filenames that do not depend on memory, such as an evidence ID, UTC date, domain and short page label. Keep the log separate from the captured files so it can describe missing, corrupted or replaced items too.

For team or legal workflows, record who captured, transferred, opened, copied or transformed the material and when. The required chain-of-custody procedure depends on the forum and jurisdiction; this guide is documentation practice, not legal advice or a promise of admissibility.

Cite what you actually preserved

A useful web citation usually includes the author or organisation, page title, publication or update date when supported, live URL, archived URL and your retrieval date. Add a section heading, paragraph description, page number or media timestamp so readers can find the relevant evidence.

Do not cite an archived homepage when your claim came from a linked PDF. Do not cite a search-result snippet when you opened the underlying page. If the live and archived versions differ, say which version supports the statement and preserve both when the change matters.

Recheck the bundle before relying on it

Run a short verification pass:

• can another person open the archive and locate the passage?
• does the screenshot show enough source context?
• does the saved file still match its recorded digest?
• are the live and final URLs distinguished?
• are publication time, retrieval time and archive time labelled separately?
• are gaps and transformations disclosed?

If any answer is no, repair the record whilst the source is still available or state the limitation clearly.

Preserve proportionately and lawfully

Public availability is not blanket permission to republish. Copyright, privacy, data-protection, court-order and contractual restrictions can still apply. Preserve only what your legitimate purpose requires, restrict access to sensitive material and redact personal data from working or shared copies without altering the retained original.

Under the Protection of Personal Information Act (POPIA), when collecting personal information you must inform data subjects of what is being collected, the source if not collected directly, your identity and address, the purpose and retention period, whether supply is voluntary or mandatory, consequences of non-provision, and the right to access, rectify and object to processing. Data subjects may object to processing by submitting Form 1 to the responsible party, provided they can show legitimate grounds for objection.

Do not create public archives of exposed credentials, intimate material, children's data, private contact details or information whose republication would create a foreseeable safety risk. Do not evade access controls. When a short quotation and a reliable citation support the work, do not republish an entire copyrighted page.

If you need to reconstruct older states, use how to find the earliest version of a web page. For a claim that has spread through repeated quotations, continue with how to trace a quote to its original source.

Apply this method in a research workflow

Use the DeepSearch Research overview to choose a starting point, then follow the source comparison workflow. Public guides are free; running Research requires paid access.