Two different people ask what AI detector students use. A student wants to know what everyone else runs before handing work in. A teacher wants to know what their class is running, and usually assumes the answer is something sinister.
The honest answer to both: students mostly use free consumer detectors to check their own drafts, the free tiers of tools such as GPTZero, ZeroGPT, QuillBot's AI detector and ours, because being wrongly flagged is a real fear and they have no way to see what their institution will see. That is a different activity from trying to beat a checker, and the distinction matters more than the tool names.
Not institutional tools, because students have no access to those.
The practical constraint shapes everything: a student cannot run the thing that will actually judge them. Turnitin is licensed to institutions, not individuals, and its AI reading is shown to staff rather than to the person who submitted. So a student cannot see their own Turnitin AI score before submitting, and in many cases not after either.
What is left is the free consumer tier of whatever detector comes up in a search. In practice that means: whatever is free, whatever needs no account, and whatever will take a whole essay without asking for a card. Ours is one of several, and we run with no account at 3 scans a day up to 1,000 words.
Two consequences worth understanding, whichever side of this you are on. Students are reading a different instrument from the one that will assess them, so a clean consumer score is not a clean institutional score and never was. And detectors disagree with each other substantially on the same passage, so the number a student sees is one reading, not the reading.
The motive matters, because it changes what the behaviour tells you.
The common assumption is that a student scanning their own essay is checking whether they got away with something. Sometimes that is true. More often the motive is the opposite, and it is worth taking seriously: they wrote it, and they are frightened of being accused anyway.
That fear is rational. Detectors produce false positives on genuinely human writing, and the error is not evenly distributed: writers working in English as a second language are flagged more often than native speakers producing comparable work. A student who knows that, and knows an accusation is difficult to disprove, has a sensible reason to want a look before submitting. The evidence on false positives is here.
Formal academic registers make it worse. Uniform sentence shapes, heavy hedging, stock transitions and a template to follow are exactly the properties detectors read as machine-made, which is why human-written literature reviews and methods sections score high. A careful student writing to a rubric is producing the kind of prose most likely to be misread.
A student who pre-checks is not thereby guilty of anything, and treating a pre-check as evidence would be a mistake. What it does tell you is that your students do not trust the process to distinguish them from a model, which is worth knowing on its own. If that is the situation, saying out loud how you handle a flag is more useful than any tool.
Three things, and the first is the one students most often get wrong.
It does not predict the institutional verdict. No consumer detector can see what Turnitin or an institutional tool will report, because those services do not expose their readings to outside products. A low score on a free checker is reassuring and is not a clearance, and treating it as one is the most common mistake here.
It does not make a flag defensible. If work is questioned, the thing that answers the question is a record of how it was written: dated drafts, version history, notes, the paragraph that got restructured twice. A screenshot of a detector score is not evidence of authorship. What actually helps is here.
It does not justify rewriting your own work to chase a number. This is the trap. A student who scores high on their own honest writing is looking at an instrument error, and editing their voice out to satisfy it makes the writing worse and the record murkier. If the work is yours, the correct response to a high score is to keep the drafts, not to launder the prose.
Read which sentences drive the reading rather than the headline percentage. That tells you something actionable about your own writing: a run of same-shaped sentences, stacked connectives, a paragraph of hedged abstraction with no specifics in it. Those are worth fixing because they are worth fixing, not because of the score.
Scans and rewrites share one daily count, so this row is the one that applies.
| Free | Starter | Pro | |
|---|---|---|---|
| Price | $0 | $9.99/mo · $3.99 yearly | $19.99/mo · $7.99 yearly |
| Actions a day (shared) | 3 | 20 | 100 |
| Words per scan | 1,500 | 5,000 | 7,500 |
| Words per humanization | 260 | 5,000 | 10,000 |
| Detection a month | No cap | No cap | No cap |
| Humanizer a month | No cap | 40,000 words | 80,000 words |
One action is one scan, one file upload or one rewrite, from a single shared daily count. That daily number is the gate that matters: there is no monthly ceiling on any plan, so what separates the tiers is how often you can work and how long a single piece can be, not how much you get through in a month.
Read the free row together rather than line by line. No monthly cap sounds unlimited and is not: three actions a day at up to 260 words a rewrite is roughly 780 rewritten words in a day. It is a real allowance for checking and fixing specific passages, and it is not a plan for processing a backlog.
Current prices, the Business and Enterprise tiers and regional pricing are on the plans page, which is the page we keep up to date.
Mostly the free tier of whatever consumer detector a search turns up, such as GPTZero, ZeroGPT, QuillBot's AI detector or ours, run on their own drafts before submitting. The reason is structural rather than a preference: institutional tools like Turnitin are licensed to institutions, not individuals, and the AI reading is shown to staff rather than to the student. So a student cannot run the thing that will actually judge them, and uses what is free and needs no account instead. Ours is one of several, at 3 scans a day up to 1,000 words with no signup.
Generally no. Turnitin is licensed to the institution and its AI indicator is surfaced to staff, so a student usually cannot see it before submitting and in many cases not afterwards either. That gap is the whole reason consumer pre-checking exists. It also means a clean score on a free detector is not a clean score on the institutional one, because they are different instruments.
Checking your own writing is not the same as trying to beat a checker, and the usual motive is the opposite of the one assumed: students who wrote the work themselves are frightened of being accused anyway. Whether any particular use is permitted is set by your institution rather than by us, and policies differ, so the handbook is the authority. What a pre-check cannot do is make a flag defensible, because that takes a record of how the work was written.
Most likely because of how formal academic prose is built. Uniform sentence shapes, heavy hedging, stock transitions and a template to follow are exactly the properties detectors read as machine-generated, which is why human-written literature reviews and methods sections routinely score high. The error is also not evenly spread: writers working in English as a second language are flagged more often than native speakers producing comparable work. A high score on your own honest writing is a fact about the instrument.
If the work is yours, no. You would be editing your own voice out to satisfy an instrument that is wrong about you, which makes the writing worse, and a heavily rewritten draft also damages the version history that would have defended you. Keep the drafts instead. The one useful thing to take from a scan is which sentences drive the reading, because a run of same-shaped sentences or a paragraph of hedged abstraction with no specifics is worth fixing on its own merits.
Most colleges that check for AI use a tool licensed to the institution and built into their submission system, most often Turnitin, and the reading goes to staff rather than to the student. That is a different instrument from the free checkers students run, which is why the two scores often disagree. What colleges actually run, and how they use it, is covered in our guide to whether colleges check for AI.
Not one of their own. Canvas and Schoology are learning platforms; where a school checks work for AI inside them, it is through a connected tool such as Turnitin, and the reading comes from that tool and is shown to staff. So the question to ask is which detector your school has connected, not what the platform does.
Not much on its own, and certainly not guilt. A pre-check mainly indicates that students do not trust the process to tell them apart from a model, which is a reasonable worry given how false positives behave. The more useful response is to state plainly how a flag is handled in your class, since the fear comes from not knowing. A detector result is a reason to look more closely and ask for drafts, never a basis for an accusation on its own.
The institution side: what gets run on submitted work.
Read the guide →Why honest academic writing scores high, and who it hits hardest.
Read the evidence →The record that answers an accusation, which a score cannot.
See the steps →Three scans a day at up to 1,000 words, no account.
Open the detector →Three scans a day at up to 1,000 words, no account. The per-sentence view is the part worth your attention.