Verification Is Not Proofreading

A draft can be clean and still be wrong. Proofreading catches the sentence that reads badly. It cannot catch the citation pointing at a retracted paper, the link that died between drafting and publication, or the claim attributed to an author who never made it. Those are different failures with different tools, and conflating them is how careful people end up publishing things they have to correct later.

Four of the five tools below check sources. One checks prose. That ratio is deliberate. A clumsy sentence embarrasses you briefly. A citation that does not survive scrutiny follows the piece around.

The Retraction Watch Database

A free, searchable record of retracted papers, maintained by the organization that has tracked retractions since 2010. You can query it directly, and several reference managers now check against it automatically and flag affected entries in your library.

This prevents the most avoidable error in scholarship: citing a paper that has been withdrawn. Retraction notices propagate unevenly. A file you downloaded two years ago carries no notice. A copy sitting on a personal page or a repository mirror may never be updated at all. The paper reads exactly as it did the day you saved it. Checking costs seconds, and the cost of not checking is the kind of error a reviewer finds and you do not.

PubPeer

A platform for post-publication commentary, where researchers raise concerns about specific papers — image duplication, statistical problems, data that does not reconcile.

Retraction is a binary, and a slow one. A great deal of work sits in the space between published and withdrawn, carrying serious public criticism that has never been resolved. PubPeer is where that space is visible. The caution is that comment quality varies and some are wrong or motivated, so treat a thread as a reason to read more carefully rather than as a verdict. But if a paper carrying weight in your argument has a thread on it, you want to find that yourself rather than hear it from a reader.

Unpaywall

Locates legal open-access versions of paywalled articles — author manuscripts in institutional repositories, publisher open editions, preprints — by checking a DOI against a large index of repositories.

This does two things. It gets you to the full text, which matters because reading an abstract is not reading a paper and the difference shows in what you write. And it lets you cite something a reader can actually open. A citation resolving to a paywall is a claim your reader has to take on faith. When a legal open version exists, pointing at it is a small courtesy that materially changes whether anyone can check your work, which is the entire purpose of citing in the first place.

The Wayback Machine

The Internet Archive's capture service. You can request a snapshot of a page on demand and receive a permanent, timestamped address for what it said at the moment you looked at it.

This solves a problem every practitioner-facing writer eventually has. You cite a policy page, a program description, an institutional report. Two years later the site is redesigned, the page is gone, and your citation resolves to a homepage redirect. The content was real. You simply cannot prove it anymore. Capturing the page at the moment you cite it costs about fifteen seconds and converts a fragile reference into a durable one. I now do this for every source that is not a journal article, and I wish I had started years earlier.

LanguageTool

A rule-based grammar, style, and consistency checker. Open source, and it can be self-hosted if you would rather not hand an unpublished draft to someone else's server.

This is the prose half, and its real value is not grammar. It is consistency across a long document — a term hyphenated on page two and not on page nine, a name spelled two ways, a tense that slips inside a paragraph. Those are precisely the errors your own eye skips, because you know what the sentence is supposed to say and you read what you meant. A checker does not know what you meant, which is exactly why it catches them. It will also flag things that are fine, and ignoring it confidently is part of using it well.

The order matters

Check the sources before you polish the sentences. This sounds obvious and almost nobody does it, because polishing is pleasant and verification is not. But a paragraph you have rewritten four times is a paragraph you have become reluctant to cut, and if the citation underneath it does not hold, cutting is what the situation requires. Verify first, while the prose is still cheap to lose.

None of this comes from being suspicious of the literature. It comes from the ordinary rate at which things move, break, and get corrected between the day you find a source and the day someone reads what you wrote about it. That interval is where the errors live. These tools are just ways of looking into it before someone else does.