Only accepted work is recordedAboutGuidesContactSign in
CitePepCitePep
Guides/Accuracy/How to spot common factual slips
Accuracy5 min read · Intermediate · Updated Sep 2026

How to spot common factual slips

The five categories that account for almost every correction we accept, with their real frequencies and the check for each.

Between 1 July and 6 October 2026 reviewers accepted 1,753 corrections across 1,700 articles on CitePep. 1,249 of them carry a category, and the distribution is narrow: five categories cover almost everything.

Wrong on the facts, 738 corrections or 59.1%. Source does not carry the claim, 529 or 42.4%. Outdated, 123 or 9.9%. Missing context, 110 or 8.8%. Wording outruns the evidence, 58 or 4.6%. The percentages exceed 100 because 331 corrections carried more than one category, usually in a single sentence.

Note what is absent. A fabricated quote appears about 3 times in 1,249, an invented threshold 3 times, a finding reported backwards 3 times. Invention is rare; being wrong about a year is routine.

The patterns at a glance

PatternHow to spot itThe quick check
Aging statistic (9.9%)A number repeated long after its surveyFind the date on the figure, then look for a newer one
Wording outruns evidence (4.6%)“Proves” where the source says “suggests”Ask what evidence earns the word you used
Source does not carry it (42.4%)A real citation that says less than the sentenceOpen the source and read it, do not just follow the link
Correlation read as cause“X causes Y” from two things moving togetherWrite “associated with” until the source shows cause
Wrong on the facts (59.1%)A wrong year, figure, statute or attributionRedo the arithmetic and check the units and base

The aging statistic

123 corrections, 9.9%. A figure accurate at publication gets quoted forward until the date comes off it. One article dated a Shopify meeting-culture change to 2021 where its source says January 2023. Another quoted a KfW StartGeld maximum loan at the cap it held before December 2025. A third put an event's attendance at 165,000 when the 2025 edition drew more than 180,000 from 171 countries.

The check: find the date on the figure, ask whether something newer exists, and put that date in the text so the reader can weigh it too.

Stronger-than-evidence wording

58 corrections, 4.6%, and the smallest category. One study “suggests”, and in the retelling it becomes “studies prove”. The evidence is often fine; the sentence has outrun it.

The fix is usually scoping rather than cutting. One correction narrowed an article's JWST framing for overstating the implications and left the surrounding astronomy alone, because the rest was already correct. Another stated the 150-coconut-deaths-per-year figure as established fact; that one is a myth with no rigorous basis, so it went.

i

A useful test: for any strong word, “proves,” “always,” “guarantees,” ask “could I show the evidence that earns this word?” If not, reach for the precise one instead.

The quote in search of a source

Quotes drift, get reassigned to more quotable names, and shed the context that gave them meaning. The failure mode is not usually invention: of 1,249 corrections, a fabricated quote appears about 3 times, while unverifiable attribution appears constantly.

Real examples from the log. A quote credited to an anonymous “cloud engineering hiring manager at a mid-size fintech company” was removed: not false, just nothing behind it. An article attributed a line to Walter Pauk that could not be verified. Another presented an invented practitioner, “Maria Gonzalez”, as a source. One cited a book title, “Mind Map Mastery”, that did not match the reference the article itself listed.

The check: trace the quote to a transcript, a documented speech or the actual book before relying on it. Untraceable means folklore, not fact.

Correlation read as cause

Two things move together and the sentence claims one caused the other. A third factor may drive both, or the direction may run the other way. This one survives review easily because it is technically sourced: the correlation is real and the causal leap belongs to the writer.

Write “associated with” unless the source establishes cause. It costs one word.

The off-by-a-scale figure

Part of the largest category, the 738 corrections that were simply wrong. Millions written as billions, a percentage taken of the wrong base, a local number presented as national.

Redo the arithmetic rather than eyeballing it. One article reported a rise from about 1 billion dollars to 9.7 billion as a 970% overrun, where it is 870%. Another stated that an organiser “declared 71% of 130,000 attendees in 2025 as decision-makers”, implying roughly 92,000 qualified buyers: the organiser's own reporting put attendance near 127,000, and nothing supported the 71% at all. A precise-sounding percentage had been layered onto an approximate headline figure.

A surprising number is a prompt to recheck the source, the units and the base.

The habits that keep content accurate

Six habits, ordered by how much they catch:

Build these into every piece

  • Open every source as you cite it and confirm it carries the sentence, which is the 42.4% case
  • Put the date beside every figure, which makes the 9.9% case visible to you and the reader
  • Trace each quote to a transcript or the actual book before relying on it
  • Keep causal language for causal evidence, and write “associated with” otherwise
  • Redo the arithmetic on every derived figure, percentages and totals included
  • Have someone with authority in the topic read it, which catches what the other five miss

The last one does the most work, because a writer reads what they meant rather than what they wrote. Every one of the 1,753 corrections here was found by someone who had not written the page.

One further step worth taking: after any correction, search the page for the wrong figure rather than re-reading it. 595 of those 1,753 found a second stale copy in a bibliography, an FAQ answer or the structured data, where the body had already been fixed. See how to review an article for accuracy.

Common questions

Which slip is the most common?

Being plainly wrong, 738 of 1,249 categorised corrections, followed by a source that does not carry its claim at 529. The aging statistic and over-strong wording are the two least common of the five, at 9.9% and 4.6%. The intuition that stale figures dominate is wrong: wrong years and unsupported citations do.

How do I catch slips in my own writing?

Mostly you cannot, because you read what you intended. The mechanical checks are the exception: opening each source, redoing each calculation and searching for a figure are all things you can do to your own draft. Judgement calls need another reader, see how to run an editorial review.

Is “associated with” just weaker writing?

It is more precise. If the evidence shows an association then “associated with” is the accurate claim and “causes” is an overstatement, which a knowledgeable reader spots immediately.

How do I check a statistic I really want to use?

Four checks: the primary source exists, the figure matches it, the date is current, and the scope fits your use. All four held is the bar; three is where the 529 unsupported citations came from. See how to cite sources properly.

Was this guide useful?