Paste 65 words or more into the checker on this page and you get a number from 0 to 100 plus one of three readings: Reads as AI-generated, Borderline, or Reads as human-written. A check costs one credit whatever the length, and the free allowance is 10 credits a day per network, so that is ten full checks a day. There is no account, no login, no card and no email address, and nothing to verify before the first run. Sitting outside that allowance is a second set of tools that run entirely inside your browser and carry no daily limit at all: a proofreader, a word counter, a readability checker, an AI vocabulary checker, a plagiarism and originality check, and a signed drafting record. Free here means two different things, and the difference is worth knowing before you spend anything.
Every run returns the same two things. A score from 0 to 100, where higher means the text carries more of the surface features these systems respond to, and a plain reading of that score in one of three bands. The band is the part to act on. Reads as human-written means those features are largely absent. Reads as AI-generated means they are present in quantity. Borderline means the reading is genuinely unresolved, and the honest way to treat it is as neither result rather than as a pass.
The floor is 65 words because a reading taken from less text than that is close to noise. Sentence length variation, phrase repetition and predictability all need a stretch of prose before they mean anything. If you are two sentences short, include the paragraph around the passage rather than checking the fragment on its own.
Price does not move with length. A check is one credit whether you paste 70 words or 2,000, because it is one pass either way. A single run takes up to 12,000 characters, roughly 2,000 words, so anything longer goes in at section size and each section still costs one credit. That is the most useful thing about how this is priced: measuring is close to free, so there is no reason to guess.
Two tools here cost credits, because both of them run work on a server. The checker at one credit a run, and the humanizer at one credit per 50 words. Ten credits a day covers ten checks, or 500 words of rewriting, or any mix of the two. The allowance refills the next day and is counted per network, not per person, so a shared office or campus connection draws on one pool.
Everything else runs locally in the page you are reading. Your text never leaves the browser for those, which is also why they can be unlimited: there is no server cost to ration. Use them as often as you like, on drafts of any length, without touching your ten credits.
A score describes style, not authorship. Text carries no record of who typed it, so nothing inside a document can report where it came from. Every detector, this one included, is reading how the sentences behave and inferring from that. A high number is a description of the writing, not evidence about the writer.
It is also not a forecast of the number another product will show you. Different model, different training data, thresholds chosen by whoever installed it, retrained on its own schedule. Two detectors can read the same paragraph and disagree. Treat the reading you get here as a reading of your draft.
That has a practical answer. A score tells you where you stand; it does not tell you which sentences did it. For that, run the AI vocabulary checker over the same text. It marks the specific word or phrase instead of scoring the document, it runs in your browser, and it does not touch your allowance. The AI writing tells reference sets out each pattern with an example, so you can see why a phrase gets marked and what to put in its place. Once you can see where the marks gather, you know which two paragraphs are worth editing and which ones to leave alone.
Most people burn their allowance by checking the same unchanged draft repeatedly. The useful pattern is check, edit, check again, where the second run is measuring an actual change rather than confirming the first one.
For a long document, split at your own section breaks rather than mid-argument, and run the parts one after another. A run takes up to 12,000 characters, about 2,000 words, so a 5,000-word chapter is three runs and three credits out of the ten. Leave out the cover page, the reference list and long block quotes. Quoted passages are somebody else's sentences and get scored like any other prose, which muddies the reading of the writing you are responsible for.
A run you cannot afford is refused before anything is deducted. You are told what it costs and what you have left, and your balance is exactly where it was, so nothing is lost by trying.
This happens, and not at random. Detectors respond to prose that is clean, even and correct, which is a description of careful writing rather than a fault in it. Stanford HAI researchers (Liang et al., 2023) ran TOEFL essays written by non-native English speakers through seven detectors and found 61% falsely flagged as AI-generated. Every one of those essays had a human author.
Do not start rewriting in a panic. Run the AI vocabulary checker over the same text first and see what actually gets marked. If the marks sit in one or two paragraphs, you have an ordinary editing job. If almost nothing is marked and the score stays high, the cause is usually register: formal, orderly, template-shaped prose, often taught. Rewriting your voice to please a classifier is a poor trade.
The stronger move is a record. Draft somewhere that keeps version history and leave your outlines and earlier files where they are, because copying text into a fresh document throws that history away. The proof-of-human-writing tool builds a signed record of your own drafting as it happens, and the verify page checks the signature on one. It works on the essay you are about to write rather than the one already submitted, which is the reason to start it early.
If the check says the draft reads as machine-written and the marked phrases confirm it, the humanizer rewrites every sentence you give it in one click. One run takes 12,000 characters, roughly 2,000 words, with a floor of 25 words. It is priced by length at one credit per 50 words, so ten credits is 500 words of rewriting. Spend one on a check first and you have nine left, which is 450 words.
Rewrite paragraphs rather than whole documents. Send everything through and the sentences you were pleased with come back changed too. Send the sections that carried the score and the writing that already worked stays yours. Then check the revised version, so the reading you act on belongs to the text you are actually submitting.
On permission, once and briefly. Rewriting a draft you were allowed to generate is ordinary editing. Where AI was not permitted, that rule covers the draft however it is edited afterwards. The policy that binds you is the one in your own handbook, so read that rather than a general rule from any website.
The edits worth making by hand cost no credits at all:
Free every day. Ten credits a day per network, refilling the next day, with a check costing one credit whatever the length. That is ten full checks a day. There is no account, no login, no card and no email address, and nothing to verify. The proofreader, word counter, readability checker, AI vocabulary checker, originality check and signed drafting record run in your browser and have no limit at all.
The minimum is 65 words and a single run takes up to 12,000 characters, which is roughly 2,000 words. There is no cost penalty for using all of it, since a check is one credit whatever you paste. For anything longer, split at your own section breaks and run the parts one after another, which also gives you a clearer reading of each section.
Higher means the text carries more of the surface features these systems respond to: predictable phrasing, flat sentence rhythm, stock constructions. The three-band reading is the part to act on. Human-written means those features are largely absent, AI-generated means they are present in quantity, and borderline means the reading is unresolved and should be treated as neither result.
No. That is a different model, with thresholds somebody local chose and its own retraining schedule, and detectors can disagree about the same paragraph. What you get here is a reading of how your draft looks on the features these systems score, plus free browser tools that mark the specific words and phrases doing it.
Not for the browser tools. The proofreader, word counter, readability checker, AI vocabulary checker, originality check and drafting record all run locally in the page, and nothing is sent anywhere. The checker and the humanizer run on a server because the work needs it, so those two send the text you paste. Neither one asks for an account, so there is no profile for anything to attach to.
Because the work scales differently. A check is a single pass over the text, and that pass costs the same whether the passage is 70 words or 2,000. Generating a rewrite takes time proportional to length, so it is priced at one credit per 50 words, rounded up to the next whole credit on each run, with a 25-word floor on the input.
Ten checks, or 500 words of rewriting, or a mix. A common split is one credit on a check, the free AI vocabulary checker to find where the problems sit, then nine credits rewriting the worst 450 words and hand-editing the rest with the unlimited browser tools.