How to Use This Tool
Paste your writing. Passive constructions are split into two lists, because they are not the same problem.
Passive voice is not an error
Style guides and word processors have spent decades treating it as one, and they are wrong. Passive is the right choice whenever the actor is unknown, irrelevant or obvious:
- The bill was passed in 1998. Nobody wants the roll call.
- The patient was admitted at 3 a.m. Which nurse is not the point.
- The species was first described in 1847. Naming the describer would derail the sentence.
Rewriting those actively makes them worse. So the useful question is not “is this passive” but “has the actor disappeared, and does that matter here?”
The distinction that does the work
With an agent — the report was written by the committee. Everything you need is still in the sentence; the passive has only changed the order. Sometimes the active version reads better, sometimes the passive keeps the topic in the subject position where it belongs. This is a preference.
Without an agent — mistakes were made. The person responsible is no longer in the sentence at all. That is legitimate when nobody knows who, and it is the oldest evasion in the language when somebody does. Only you can tell which, but you can only tell if the tool shows you where to look.
How the detection works, and where it fails
It looks for a form of be — is, are, was, were, been, being, be — optionally followed by an adverb, then a past participle. Regular participles end in -ed; irregular ones come from a built-in list (written, taken, given, shown and so on). A common list of predicate adjectives is excluded, which is what stops was tired and is important being flagged.
Without a parser it cannot be perfect, and the failures run in both directions:
- Missed: passives built with get rather than be — he got fired — and participles far from their auxiliary.
- Over-flagged: a participle used as an adjective in an unusual position, and -ed words that are simply past tense verbs sitting next to a form of be.
The lists are tuned to under-flag rather than over-flag, because a checker that cries wolf is one people stop reading.
A number worth ignoring
The percentage is shown because people ask for it, not because there is a correct value. Some publishers say 10%, some say 20%. Academic and technical writing legitimately runs higher, and a narrative can sit near zero without trying. Read the agentless list; the percentage is decoration.
