Passive Voice Detector — Find Weak Sentences in Body Copy

Passive voice is a choice, not a defect

In the passive, the thing being acted on becomes the subject and the actor moves to the back behind "by" — or disappears altogether.

Active: The marketing team wrote the report.
Passive: The report was written by the marketing team.
Passive, actor gone: The report was written.

That third line is where the real problem lives. English needs the passive and uses it constantly; what causes damage is not the construction but the vanishing actor. "Mistakes were made" is the famous case precisely because the grammar is unremarkable and the missing subject is doing all the work.

What it costs when it is the wrong choice

• Extra words. A form of "to be" plus a participle, where one verb would have done.
• A delayed verb. The reader reaches "was" or "were" before reaching the action, so the sentence takes longer to land.
• An optional actor. Once the actor is grammatically optional, it starts going missing — sometimes carelessly, sometimes because nobody wanted to name it.

When the passive is right

These are genuine uses, not concessions:

• The actor is unknown. "The package was damaged in transit" — you do not know who did it, and inventing an actor would be worse than omitting one.
• The actor is irrelevant. "The samples were stored at four degrees" — which technician put them in the fridge does not matter to anyone reading a method.
• The affected thing is the topic. "Shakespeare was born in 1564" is about Shakespeare, so Shakespeare goes first. Rewriting it actively puts his mother in a sentence that is not about her.
• Convention expects it. Scientific methods sections and much legal drafting are written this way, and an actively-voiced lab report reads as though it does not know the genre.

Rewriting any of those to the active voice makes them worse. That is why the output is a list to read, not a list to fix.

How the detection works, and what it misses

It looks for a form of to be — am, is, are, was, were, be, been, being — followed immediately by a word of more than three letters ending in ed, en, wn, gn, rn, de, ne or ze. Each sentence is flagged at most once, and the panel reports the count and the ratio alongside the flagged sentences themselves.

The blind spot is specific and worth knowing: irregular participles that end some other way are missed entirely. "The deadline was missed" is caught; "the invoice was sent", "the file was lost", "the fence was built", "the shares were sold", "the error was found", "the meeting was held" and "the story was told" are all passive and all invisible to it. So the reported ratio is a floor, not a total. Common irregulars that are caught include written, known, broken, taken, given, shown, made, done and born, since those happen to end in the right pairs.

In the other direction it excludes a list of words that end like participles but are adjectives or nouns — open, fine, wide, online, modern, machine and similar — because without that list ordinary active sentences such as "the form is open" came back flagged. The le ending was dropped from the test altogether, since English has effectively no past participles ending in a bare -le and it only ever matched words like able, simple, possible and reliable.

Using the output

Take each flagged sentence and ask one question: who did this?

• If the actor is missing and you know it, try putting it in front and see whether the sentence improves.
• If the actor is present but trailing, decide whether it belongs at the front here. Sometimes the current order is right because the object is the topic.
• If the passive is deliberate, leave it, and note that you decided.

The on-screen guidance suggests keeping body copy under roughly ten percent passive. Emphasis on roughly. A draft at five percent and a draft at twelve can both read well, and a draft forced to five percent by mechanically inverting every sentence usually reads worse than the one it replaced. Never rewrite to move a number.

For the other reading on the same draft, the readability score measures sentence length and word length instead — a different problem, and one this tool cannot see.