How to Fact-Check Your Novel
Readers will forgive a dragon. They will not forgive a ZIP code that didn't exist yet. Here is a practical method for catching the errors that quietly cost you a reader's trust.
Why fiction still has to be accurate
Invented worlds get a pass. The real one doesn't. The moment a reader spots something they know is wrong — a revolver with too many shots, a Tuesday that was actually a Sunday, a drug that wasn't approved until years later — a small door closes. They stop falling and start checking. And once a reader is checking, every other detail is suspect too.
The goal of fact-checking a novel isn't pedantry. It's protecting the dream. A book that gets its real-world details right earns the authority to be believed about everything it invents.
What actually needs checking
You don't have to verify every sentence. Concentrate on the five categories where errors actually happen:
- Period and historical detail. When did the thing exist? Zip codes arrived in 1963. The Boeing 727 first flew in 1963. "Ms." wasn't in common print use until the early 1970s. Anachronisms are the most common — and most catchable — error in historical fiction.
- Technical and procedural accuracy. How does an arraignment actually run? How long does rigor mortis take? What can a nurse legally do that a doctor can't? Readers with firsthand knowledge are the ones who write reviews.
- Geography and travel. Can your character really drive that distance in that time? Does the river run the direction you say? Is the neighborhood where you put it?
- Real events on your timeline. If your story is set against real history, the real history has to line up — what was on the news, who was president, what the weather did.
- Your own internal facts. The most frequent errors aren't about the world; they're about your book. A character's eye color, a house's layout, the age you gave someone in chapter two. (That's continuity, and it deserves its own pass.)
A method that works
Fact-checking goes wrong when it's a vague intention to "look things up later." Make it a discrete pass with four steps:
- 1. Flag the claims. Read for assertions about the real world — dates, technologies, procedures, places, numbers. Mark each one. Most writers are shocked how many there are; a single page can carry a dozen quiet factual claims.
- 2. Find a real source. Not your memory, and not a chatbot's memory — an actual source you can point to. Memory is exactly where anachronisms hide, because the thing feels period-appropriate.
- 3. Verify or fix. Confirm it, or change the text. When you fix, check that the fix didn't break something downstream (moving a date can ripple).
- 4. Keep the receipt. Note the source next to the fact. When an editor or a reader challenges it later, you have the answer already filed.
Where AI helps — and where it hurts
A general-purpose chatbot is a poor fact-checker for one specific reason: it will confidently tell you something is fine when it isn't, because it's drawing on the same fuzzy memory that let the error in. Worse, most writing AIs are built to write — offer to rewrite the passage, flatter the prose — which is the opposite of what checking needs.
A useful checker does three things instead: it retrieves real sources and cites them, it returns findings only — what's wrong, with evidence, not compliments — and it vets each finding before you see it, so you're not chasing false alarms. If a tool can't show you why it flagged something, treat the flag as a rumor.
How Skarvia does it
This is the exact job Skarvia's fact-checker was built for. It reads your draft, checks the real-world claims against sources, and returns defect cards — the quote, the problem, the evidence, and a citation — or one quiet line: "No issues found." Every finding first has to survive a second reviewer whose only job is to refute it, so what reaches you is worth stopping for. It never rewrites a word; the fix stays yours.