All posts

Fact-Checking with Web Archives: How to Verify What a Website Really Said

July 23, 2026

"The website used to say something different" is an easy claim to make and a hard one to prove. Screenshots are trivially fabricated, memories are unreliable, and live pages change without notice or acknowledgment. Web archives are the closest thing the internet has to a neutral witness — but only if you use them rigorously. A sloppy archive lookup can lend false confidence to a wrong conclusion just as easily as it can verify a true one.

Pin down the claim first

Before opening an archive, reduce the claim to something checkable: a specific URL, a specific statement, and a specific time window. "They changed their refund policy" becomes "example.com/refunds said 30 days in January 2024 and says 14 days now." Vague claims produce vague verifications. If the person making the claim cannot name a URL or a date range, that is worth noting before you spend an afternoon searching.

Find captures on both sides of the change

Verification means bracketing: a snapshot showing the old text before the alleged change, and one showing the new text after. The Wayback Machine's calendar view for the URL gives you the capture timeline. Pull the nearest captures on each side of the claimed change window and compare the relevant passage directly. The narrower the gap between your two snapshots, the tighter you can date the change — and the digest hashes in the CDX API can tell you exactly which captures differ without opening each one.

Beware of rendering artifacts. A passage that seems missing from a snapshot may have been loaded by JavaScript the crawler did not execute, hidden in a collapsed section, or served differently by region or device. Check the page source of the snapshot before concluding text was absent. Absence from one capture is weak evidence; absence from several captures across months, on a page that otherwise rendered completely, is much stronger.

Corroborate across archives

A single archive can have a bad capture. When the stakes are real, confirm the key snapshot in a second source — Archive.today often has independent captures, and Memento Time Travel checks multiple archives at once. Two independent archives showing the same text on overlapping dates is close to conclusive. Also look sideways: contemporaneous quotes in news coverage, social media posts, and forum threads that cite the original wording all corroborate what the page said at the time.

Arkibber fits the corroboration phase well: moving between captures, dates, and related items quickly without losing your place. Fact-checking is comparison-heavy work, and every minute spent re-finding a snapshot you already had is a minute not spent evaluating it.

Report what the evidence supports — and no more

An archive proves what the crawler saw at a URL on a date. It does not prove intent, authorship, or who saw the page. "The page was changed to conceal X" is an inference; "the page said X on March 3 and no longer said it by March 19" is a finding. State findings with their snapshot dates and archived URLs, flag the gaps (no captures between the bracketing dates, JavaScript-rendered sections you could not verify), and let the evidence carry the argument. A fact-check that names its own limits is harder to dismiss than one that overclaims.

Finally, preserve your evidence at the moment you find it. Save the snapshot URLs, download or screenshot the key captures, and archive the current live page too — the person you are checking may edit it again once questions start being asked.

How to Find Old Newspapers and Magazines on the Internet Archive