How to fact-check an article: an editor's checklist (with free tools)

This is the editor's version of our fact-checking series. For the tools that automate parts of it, see the pillar guide to AI fact checkers.

The 10-point fact-checking checklist

Print this, or copy it into your editing doc.

FACT-CHECK CHECKLIST                     Story: ____________  Checker: ______

[ ] 1. Every checkable claim is marked and listed on the fact sheet.
[ ] 2. Every person: name spelling, title, affiliation, age confirmed.
[ ] 3. Every quote: checked against the recording or notes, in context.
[ ] 4. Every number: matches the primary source; units, dates, base checked;
       derived numbers recalculated.
[ ] 5. Every date, place and sequence of events confirmed.
[ ] 6. Every study: exists, says what we say, peer-review status noted,
       not retracted.
[ ] 7. Every document and link: opens, is the original, says what we say.
[ ] 8. Every image, video and caption: origin and date verified.
[ ] 9. Implications checked: headline, framing and omissions are fair to
       the evidence.
[ ] 10. Open items resolved with writer and editor; fact sheet saved.

Types of fact-checking: pre-publication vs. post-publication

Fact-checking is the process of verifying every checkable statement in a piece, before or after it is published, against the best available source. There are two traditions:

Most small teams do a light version of pre-publication checking: the writer checks, the editor spot-checks. The checklist above is the full version, condensed so one person can run it on a feature in an afternoon.

Build a fact sheet: every claim, its source, its status

The fact sheet is the core tool. One row per claim. Start it before you check anything.

| # | Para | Claim (exact words)           | Source (doc, page, link, person + date) | Status         | Notes / fix              | Checker |
|---|------|-------------------------------|------------------------------------------|----------------|--------------------------|---------|
| 1 | 2    |                               |                                          | Confirmed      |                          |         |
| 2 | 2    |                               |                                          | Corrected      |                          |         |
| 3 | 4    |                               |                                          | Unconfirmed    |                          |         |
| 4 | 5    |                               |                                          | Cut            |                          |         |

Status values: Confirmed / Corrected / Unconfirmed / Cut. Nothing runs as "Unconfirmed" without an editor's decision.

Ask the writer for an annotated draft: each claim tagged with its source. The writer knows where every fact came from; reconstructing that later takes longer.

Checking people, titles and quotes (call-backs, recordings)

People and titles

• Spell every name against an official source: the person's own bio, an organization's staff page, a court record. Then ask the person if you can.
• Check the title as of the date in the story. People change jobs; a title that was right last year may be wrong now.
• Check ages against a date of birth, not a previous story.

Quotes

• Check every quote against the recording or contemporaneous notes, word for word.
• Read the sentences before and after. A quote can be accurate and misleading.
• Confirm who said it. In group interviews and transcripts, speakers get mixed up.
• Call-backs: checkers often contact sources to confirm facts attributed to them. Confirm facts, not wording, unless your outlet's policy says otherwise, and don't send a source the full draft.
• If a quote came from an AI tool or an unsourced web page, treat it as fake until you find the primary source. Our guide to fact-checking AI-generated content covers why.

Checking numbers and statistics (base rates, percentages, AP numbers style)

Numbers fail in predictable ways. Check each against this list:

The examples above are illustrative, not real data.

Style. If your outlet uses AP style, numbers have their own rules: spell out one through nine, figures for 10 and up, "percent" spelled out, and exceptions for ages, money and dimensions. They're all in our AP style numbers, dates and percent cheat sheet.

Checking scientific and medical claims (peer review, retractions, preprints)

Science claims need four checks:

1. Find the paper. Not the press release, not the news story. Confirm the DOI resolves to it at https://doi.org/.
2. Check status. Peer-reviewed journal, or preprint? A preprint can be cited, but the story should say it hasn't been peer-reviewed.
3. Check for retraction. Search the title or DOI in the Retraction Watch Database, which Crossref acquired and made free in 2023.[1] Also check the journal page for corrections or expressions of concern.
4. Check the claim against the results. Is it the paper's main finding or a side result? In people or in mice? Correlation or causation? How large was the effect, and how large the sample?

For health claims, also check whether the finding matches the wider evidence. One study rarely settles a question, and a story that says otherwise needs a strong reason.

Free fact checking tools

Two cautions. A reverse image search that finds no earlier copy proves nothing. And a fact check from another outlet is a lead, not a verdict: read their sources.

Using comments and suggestions as your fact-check log

Many small teams skip the fact sheet because it lives in a separate spreadsheet that falls out of date. Keeping the log in the draft itself fixes that.

In WriteOrbit, run AI review on the draft: claim checks leave a comment on each claim with the source matched, and the source-quality check flags citations to retracted papers, preprints and DOIs that don't match the paper. The checker replies to each comment with the source and status ("Confirmed: council minutes, p. 3") and resolves it. Fixes go in as suggestions (tracked changes), so the writer sees exactly what changed and why. Version history keeps the checked version, which you'll want if a claim is challenged later.

It doesn't replace the recording, the phone call or the reverse image search. It keeps the record of them next to the words they check. See how it works.

Corrections policy: what to do after publishing

Errors get through. What matters is how fast and how openly you fix them. The SPJ Code of Ethics asks journalists to "acknowledge mistakes and correct them promptly and prominently."[5]

A workable policy for a small team:

1. Log every error report with date, who reported it and what they said.
2. Check it with the same rigor as the original fact-check. Pull the fact sheet.
3. Fix the text if it's wrong, and add a correction note saying what was wrong and what's right. Don't silently edit.
4. Match prominence to harm. A misspelled name gets a note at the end; an error that changed the story's meaning gets a note at the top, and possibly a new headline.
5. Correct everywhere it ran: social posts, newsletters, syndication partners.
6. Review the cause. Was it a source, a process gap, a tool? CNET's 2023 review of its AI-assisted explainers found errors in 41 of 77 stories; it corrected them with editor's notes and paused the tool.[6]

Correction note template:

Correction, [Month Day, Year]: An earlier version of this article [misstated /
misidentified / incorrectly reported] [what]. [Correct fact].

FAQ

What are the steps to fact-check an article?

Mark every checkable claim, build a fact sheet listing each claim and its source, then check people and titles, quotes, numbers, dates, scientific claims, images and links against primary sources. Check what the piece implies as well as what it says. Resolve every open item with the writer and editor, and keep the fact sheet in case a claim is challenged after publication.

What is the difference between pre-publication and post-publication fact-checking?

Pre-publication fact-checking verifies every claim in a piece before it runs, the model used by magazines and many investigative teams. Post-publication fact-checking assesses claims that are already public, such as a politician's statement or a viral post, and publishes a verdict. Newsrooms also check their own published work after the fact when a reader or source flags an error.

How do you fact-check a source?

Ask who created it, how they know, and whether it is the original. Trace any claim back to its primary source: the dataset, document, study or recording. Check the author's expertise and possible conflicts of interest. For studies, confirm the paper exists, is peer-reviewed, and has not been retracted. For web pages, check the date and archived versions.

How do you fact-check statistics?

Find the number in the original dataset or report, not in coverage of it. Confirm the unit, time period, population and definition. Recalculate any percentage, change or total yourself. Check whether a percent change is being confused with a percentage-point change, and whether a small base makes a big percentage meaningless. Note the margin of error for surveys.

What free fact checking tools do professionals use?

Common free tools include Google Fact Check Explorer for published fact checks, the Wayback Machine for old versions of web pages, TinEye and the InVID-WeVerify plugin for images and video, the Retraction Watch Database for retracted studies, and doi.org for checking citations. Free AI review tools can help flag claims, but each result still needs a human check.

Where to go next

To speed up the first steps with AI, see the pillar guide to AI fact checkers and the claim-by-claim workflow. If the draft started as chatbot output, use the seven-step check for AI-generated content. For the reporting side, see how to write a news article.

Sources

1. Retraction Watch (2023). The Retraction Watch Database becomes completely open, and RW becomes far more sustainable. https://retractionwatch.com/2023/09/12/the-retraction-watch-database-becomes-completely-open-and-rw-becomes-far-more-sustainable
2. Google News Initiative. Google Fact Check Tools. https://newsinitiative.withgoogle.com/resources/lessons/google-fact-check-tools/
3. WITNESS Media Lab (2019). How to use InVID-WeVerify. https://citizenevidence.org/2019/12/11/how-to-use-invid-the-swiss-army-knife-of-digital-verification/
4. Bellingcat Online Investigations Toolkit. TinEye. https://bellingcat.gitbook.io/toolkit/more/all-tools/tineye
5. Society of Professional Journalists. SPJ Code of Ethics. https://www.spj.org/ethicscode.asp
6. Engadget (2023). CNET had to correct most of its AI-written articles. https://www.engadget.com/cnet-corrected-41-of-its-77-ai-written-articles-201519489.html
7. Crossref. Retraction Watch data documentation. https://www.crossref.org/documentation/retrieve-metadata/retraction-watch/