Reviewing an article for accuracy is a craft with a clear method: isolate each checkable claim, trace it to a real source, judge whether the source actually supports it, and record what you found. The steps are simple; the value is in doing them rigorously and in knowing the judgment calls along the way. This guide walks through the whole process, plus the harder cases, how to handle conflicting sources, claims you can’t fully confirm, and how to keep the work efficient.
Start by isolating every checkable claim
Before you verify anything, read the article and pull out every statement that can be true or false: a fact, a figure, a date, a name, a direct quote, a cause-and-effect claim. This sounds mechanical, but it’s the step that makes the rest fast and complete. A claim you never isolated is a claim you’ll never check, and those are exactly the ones that surface later.
The key skill here is separating claims from opinions and framing. An opinion (“the policy was a bold move”) carries no factual burden. A claim (“the policy raised costs by 12% in its first year”) does. Many sentences blend both, and your job is to lift out the checkable part.
“The new policy was unpopular and poorly designed.”
“The policy raised costs by 12% in its first year” (the verifiable fact, separated from the judgment).
Work through the whole piece this way and you’ll usually end up with a list of ten to thirty discrete claims for a typical article. That list is your checklist, and tracking it is what lets you say, honestly, “I checked everything” rather than “it reads fine to me.”
Trace each claim to its primary source
For every claim, the question is the same: where did this come from, and how close is that to the original? The strongest footing is the primary source, the actual study, the dataset, the official document, the first-hand account, not a news article summarizing it, and not a blog summarizing that news article. Every hop away from the primary source is a place where a number got rounded, a caveat got dropped, or a finding got overstated.
A practical example: an article claims “remote workers are 13% more productive.” You find the figure in a magazine piece, which cites a press release, which references a working paper. Follow the chain all the way to the working paper, and you may find the 13% applies to a specific group under specific conditions, useful nuance the summaries stripped out. Citing the original both strengthens the claim and protects you from inheriting someone else’s error.
A reliable shortcut: when two independent primary sources agree on a figure, your confidence rises sharply. When you can only find one source, and everything else just re-quotes it, treat the claim as resting on that single source, and say so if it matters.
Verify figures for accuracy and currency
Numbers carry two separate risks: being wrong, and being out of date. A statistic can be perfectly accurate for 2018 and misleading in 2025. So for every figure, confirm both that it matches the source and that no newer figure has superseded it. Then note the date alongside the number, so a reader can weigh it for themselves.
Scale and units deserve special attention because the slips are easy to make and easy to catch. Millions written as billions, a percentage taken of the wrong base, a national figure presented as global, these jump out the moment you sanity-check a number against what you’d expect. If a figure seems surprisingly large or small, that surprise is a prompt to recheck the source, not a finding to publish.
Check quotes word for word, and in context
Direct quotes carry two duties: the words must be exact, and the context must match what’s implied. Paraphrases drift over time, and a quote can be perfectly accurate yet still misleading if the surrounding context changes its meaning. Always check a direct quote against the original transcript or document, and confirm the speaker said it in a situation that fits how the article uses it.
This is also where attribution mistakes hide. A famous line gets reassigned to a more famous person; a sentence from a critic gets framed as the author’s own view. Tracing the quote to its real source resolves both.
The judgment calls: conflicting and unconfirmable claims
Not every claim resolves cleanly, and how you handle the messy ones is what separates a careful review from a superficial one.
When sources conflict: present what each one shows rather than silently picking a winner. If reputable sources genuinely disagree, the honest move is to reflect that, “estimates range from X to Y, depending on method”, which is both more accurate and more useful to the reader than false certainty.
When a claim can’t be confirmed either way: mark it clearly as unconfirmed. This is real information, not a gap to paper over. A claim flagged “no primary source found” tells the publisher exactly what needs a source, a hedge, or a cut, and it keeps your record honest. A review that quietly passes claims it couldn’t verify is the one that comes back to bite.
Record each claim’s status
The output of a review is a per-claim verdict: verified, corrected, or to-confirm. That record is what makes “reviewed” mean something, both to the publisher deciding what to do, and to a reader who wants to know what was actually checked. “Looks good” is not a review; a list of claims with their status is.
On CitePep, an accepted review becomes a public record of exactly which claims you checked and their status, on the page and on your profile. That turns the work you did into a credible, visible signal rather than an invisible favour.
How to keep it efficient
Rigour and speed aren’t opposites once you have a rhythm. Check the highest-stakes claims first, the statistics, the strong assertions, the load-bearing facts that the whole piece rests on, because those are where an error costs the most. Keep a running list as you go rather than re-reading the article five times. And lean on the claims that come pre-sourced; a writer who already linked a primary source has done half your work, and you simply confirm the link supports the line.
Quick checklist
- Every checkable claim isolated from opinion and framing
- Each claim traced to its primary source
- Figures confirmed accurate and current, with dates noted
- Direct quotes verified word for word and in context
- Conflicting sources reflected honestly, not silently resolved
- Unconfirmable claims marked clearly, not passed
- Every claim recorded as verified, corrected, or to-confirm
Common questions
What makes a source reliable?
A primary source, the original study, dataset, or document, gives you the strongest footing because nothing has been summarized or reinterpreted between it and you. Beyond that, prefer sources with transparent methods and a track record. When reputable sources disagree, the reliable move is to present what each shows rather than pick one.
How is reviewing for accuracy different from editing?
Editing sharpens clarity, flow, and structure; reviewing for accuracy confirms that the claims are true and sourced. They’re distinct passes and often distinct people, an excellent editor can still miss a subtly wrong statistic in an unfamiliar field. See how to run an editorial review for how the two fit together.
What do I do with a claim I genuinely can’t verify?
Flag it as unconfirmed and say why (“no primary source found”). From there it gets a source, gets softened to match what’s actually known, or gets cut. Leaving an unverifiable claim in as stated fact is exactly the kind of thing that erodes trust later.
How long should reviewing an article take?
It scales with the number and stakes of the claims, not the word count. A short piece dense with statistics can take longer than a long, lightly-claimed one. Checking the highest-stakes claims first means even a time-boxed review covers what matters most.
Can I review a topic outside my expertise?
You can verify anything you can trace to a clear primary source, but subtle, field-specific claims are best checked by someone with real authority in that subject. Be explicit about which claims you checked and which sit outside your confidence, see how to check a page for accuracy.