Why this is detection by pattern, and what that costs
An English passive is formed from a be-verb and a past participle: was built, is being reviewed, had been sent. A pattern matcher can find that shape reliably. What it cannot do is distinguish it from a be-verb followed by a participial adjective, which has exactly the same shape: was tired, is interested, was closed in the sense of being shut rather than in the sense of somebody closing it. Telling those apart requires knowing whether the word describes a state or reports an action, and for many words it requires knowing the wider context — the shop was closed is a state at 9pm and an action if the sentence continues by the landlord.
So the tool flags candidates and hands you the sentence. Confidence is graded on the evidence available: a following by phrase names an agent and is the strongest signal there is, an irregular participle such as built or written is rarely adjectival, a regular -ed form is genuinely ambiguous, and a participle on the common-adjective list is marked low. That grading is a heuristic on top of a heuristic. It ranks your reading order; it does not decide anything.
What gets flagged
| Pattern | Example | Confidence |
|---|---|---|
| be-verb + participle + by | The joists were installed by the owner | High |
| be-verb + irregular participle | The permit file was lost | Medium-high |
| be-verb + regular -ed | A load test was performed | Medium |
| be-verb + adverb + participle | The spacing was probably chosen | Medium |
| be-verb + common adjective form | I was tired | Low |
Up to three adverbs are allowed between the be-verb and the participle, which catches was not entirely rebuilt while stopping the matcher from wandering across half a sentence and pairing two unrelated words. The be-verbs recognised are am, is, are, was, were, be, been and being, so progressive passives (is being carried) and perfect passives (had been moved) both surface, the latter through the been form.
What it misses, which matters more than what it over-flags
Three real passive constructions carry no be-verb and are therefore invisible here. The get-passive — the proposal got rejected, she got promoted — is common in speech and increasingly in writing. The reduced relative clause — the report filed last week, meaning the report that was filed last week — drops the be-verb entirely. And a bare participial phrase opening a sentence — Written in haste, the memo said too much — is passive in effect with nothing to match on.
There is a fourth gap that is not really passive at all but produces the same fog: the nominalisation. The installation of the joists was carried out has taken a verb, turned it into a noun, and then needed a filler verb to make a sentence out of it. No passive detector finds those, because the passive is a symptom rather than the disease. If a paragraph feels evasive and this tool reports nothing, look for nouns ending in -tion, -ment and -ance that used to be verbs.
Passive voice is frequently the right choice
The advice to avoid the passive is a useful default badly stated as a rule. There are at least five situations where the passive is not merely acceptable but correct. When the actor is unknown: the window was broken overnight. When the actor is irrelevant: the samples were refrigerated for six hours — in a method section nobody cares which technician opened the fridge, and naming one would be noise. When the actor is obvious: the bill was passed in March. When the thing acted on is the topic of the paragraph and the passive keeps it in the subject position, which is what maintains the thread from one sentence to the next. And when the actor is being deliberately, honestly de-emphasised, as in scientific convention.
The passive is a problem when it is used to hide an actor who should be named. Mistakes were made is the canonical example and it is a moral failing rather than a grammatical one. The way to use this page well is to look at each high-confidence flag and ask who did it, then decide whether the reader needs to know. If they do and the sentence does not say, rewrite it. If they do not, leave it alone.
A note on the percentage
The summary reports what share of your sentences contain at least one candidate. Resist the urge to drive that number down for its own sake. Published, well-edited prose in most genres carries passives in something like ten to thirty percent of sentences, and the figure is much higher in scientific and legal writing where the conventions demand it. A draft that hits zero has usually been mangled: someone has invented an actor for a sentence that did not need one, or twisted a natural topic-first order into an awkward actor-first one. Use the number to notice a change between drafts, not as a target.
For a different angle on flabby prose, the filler and hedge word checker looks at intensifiers and qualifiers rather than verbs, and the sentence length analyzer catches the run-ons that passives often hide inside. If you are comparing two revisions of the same passage, text diff will show what actually changed.
Questions people ask
Why is "I was tired" flagged as passive?
Because it has the exact shape of a passive: a be-verb followed by a word ending in -ed. Nothing in the string distinguishes it from "I was hired", which is unambiguously passive. The tool marks it low confidence because "tired" is on a list of participle forms that usually function as adjectives, and there is an option to hide that whole class. Turning that option on will also hide genuine passives such as "the account was closed by the bank", which is the trade you are making.
It found nothing but the writing still feels evasive.
Look for the constructions this tool cannot see. Get-passives ("got rejected") have no be-verb. Reduced relatives ("the report filed last week") drop it. And nominalisation — turning a verb into a noun and propping it up with a filler verb, as in "an assessment was undertaken" or "the implementation of the changes occurred" — produces the same fog without a passive necessarily being present. The weak-verb list on this page catches some of the filler verbs that go with nominalisations, which is the closest signal available.
What percentage of passive sentences should I aim for?
None in particular. Published nonfiction commonly runs somewhere between ten and thirty percent of sentences containing a passive, and formal scientific writing runs much higher because the conventions of a methods section actively require it. Chasing zero produces worse prose: you end up inventing actors for sentences that did not need one, and breaking the topic thread that keeps a paragraph coherent. Track the number between your own drafts if you like, but do not treat any particular figure as a goal.
Does it handle "is being reviewed" and "had been sent"?
Yes. The be-verb list includes "being" and "been", so progressive passives and perfect passives both match on that second element. Up to three adverbs are permitted between the be-verb and the participle, which means "was not entirely rebuilt" and "is being carefully reviewed" are caught, while the matcher still refuses to pair words that are further apart than that and are probably unrelated.
Can I paste something confidential?
The scan runs entirely in this page on your own machine, and no part of the text is transmitted anywhere at any point. There is no server call in the code path and no analytics hook on the input. If you want to confirm it, load the page and then disconnect from the network; everything still works.