Wikipedia defines vandalism narrowly. It is editing in a way that is deliberately intended to obstruct or defeat the encyclopedia’s purpose. The key word is deliberately. An edit that is wrong, badly sourced or unbalanced is not vandalism if the person making it was trying to help. Wikipedia calls that a content dispute, and it handles the two completely differently.
That distinction matters more than it sounds, because the vandalism that gets talked about is almost never the vandalism that hurts anyone.
The kind that gets caught
Most vandalism is loud. Someone blanks a section, replaces a paragraph with an insult, or inserts something obviously absurd. This is the version people picture, and Wikipedia is good at catching it.
There are several overlapping defenses. Automated tools such as ClueBot NG score edits for vandalism signatures and revert the worst of them without a human being involved. Recent-changes patrollers work through the live feed of edits looking for problems. Anyone watching the article sees it in their watchlist. Crude vandalism on a well-watched article often does not survive long.
If your concern is that somebody will replace your biography with obscenity, the system mostly has that covered, and the damage from a revision that existed briefly is usually limited.
The kind that does damage
The edits that cause real problems look nothing like that. They look like ordinary editing.
A sentence gets added in the right register, in the right section, with a citation attached. The citation is real. It points at a real article. The article does not quite say what the sentence claims, or says it about a different year, or is an opinion column being cited as fact. A patroller skimming the diff sees a sourced addition in encyclopedic tone and moves on, because there are thousands of edits an hour and this one does not look wrong.
Other shapes of the same problem:
- Selective omission. Nothing false is added. Something material is removed, and the remaining text is accurate but leaves a different impression.
- Framing. “Was investigated” becomes “was investigated for fraud” where the investigation concerned something else. Both versions are short, both read plausibly, and the diff is four words.
- Weighting. A minor incident from fifteen years ago gets its own section, properly sourced, in an article where nothing else has one. Every fact is true. The article now says something the facts do not.
- Stale accuracy. A statement that was correct when written and is no longer correct. Nobody vandalized anything. The article simply stopped being true, and no one noticed.
None of these trips an automated filter, because none of them carries a vandalism signature. Several of them are not vandalism at all under Wikipedia’s definition, which is exactly why they last.
Four words, worked through
Take a biography containing this sentence: “In 2015 the company was investigated by state regulators over its billing practices, and no action was taken.”
An editor changes it to: “In 2015 the company was investigated by state regulators over its billing practices.” Eleven bytes removed. In the history list this shows as a tiny edit, smaller than a typo fix, and the edit summary says “trim”.
Everything left behind is accurate. The investigation happened, regulators ran it, billing was the subject. What has gone is the outcome, and the outcome was the only part of the sentence that was good news. A reader now finishes the paragraph knowing a company was investigated and nothing else.
No automated tool flags this. It adds nothing false, uses no unusual words, and removes fewer characters than a stray space. A patroller glancing at the diff sees a short deletion in an article they know nothing about. Under Wikipedia’s own definition this may not even be vandalism, because a good-faith editor could argue the clause was unsourced or that outcomes belong in a different section.
That is the shape of the edits that matter. Small, plausible, defensible, and expensive.
Why duration is the thing to measure
A bad edit’s damage is roughly a function of how long it is live, and that is not intuitive.
Wikipedia articles rank highly for name searches. They are scraped, mirrored across dozens of sites, and pulled into search result panels and AI assistant answers. A claim that sits in an article for a month does not stay in that article. It gets copied to mirrors that never update, cited by writers who assume Wikipedia checked, and absorbed into systems that were trained or retrieved while it was live. Reverting the original edit fixes Wikipedia. It does not fix anything downstream.
This is the practical argument for caring about speed rather than accuracy of detection. An edit caught in minutes is a non-event. The same edit caught in six weeks is a research project.
Who is actually exposed
The people most at risk have an article and no audience. Fame is close to irrelevant here, and can even help.
A heavily watched article has many editors who notice changes. An article about a mid-sized company, a regional figure, an academic or an executive may have almost nobody watching it. Wikipedia’s coverage of living people is governed by a strict policy that requires contentious material to be sourced and removed on sight if it is not, but that policy is enforced by people who are looking. On a quiet article, nobody is.
What to do about it
Three things, in order of how much they matter.
Know when it happens. Everything else depends on this, and it is the part nobody has. The page history records every change, but only if somebody goes and reads it. Here is how to read one.
Do not fix it yourself. The instinct on finding a bad edit about you is to correct it. Wikipedia’s conflict-of-interest guidance strongly discourages editing articles you are the subject of, and an account that does it is visible and traceable. The right route is the article’s talk page, or a request through the channels Wikipedia provides for subjects of biographies. We wrote about the rules here.
Treat the diff as the evidence. Before responding to anything, read the actual change rather than the edit summary describing it. The two disagree more often than you would expect.
Monitoring is the first of those and the one a service can solve. We watch the page and email you within 60 seconds of any change, with the diff in the message, so you are reading the edit rather than hearing about its consequences. See how it works, or start with the page about you.