AI-Detection.io checks one text at a time. You paste a piece of writing or upload one TXT, DOCX or PDF file, and each check takes at least 50 and at most 5,000 words. There is no bulk or batch upload, so a class set means running the checks one by one. This guide covers how to do that without turning a score into a verdict, and where to look if you need true batch scanning.
Key takeaways
- There is no batch upload here. Each check is a single text, and a signed-in account has a daily allowance, so a large set can take more than one day.
- Do not check every essay by default. Checking only the pieces that already worry you means fewer results to misread.
- Record what you checked and what you did next, so one number never stands in for a decision.
- If you need a whole set scanned in one go, an institutional tool built into your learning platform is the usual route. Check its data policy first.
Can this checker scan many essays at once?
No. Each check is one text, and the checker returns one result with highlighted passages and a confidence label for each. Checking your own text needs a free account, which is a sign-in by email link. Signed-in accounts get a larger daily allowance, and the exact figures can change, so the AI detector for teachers page and its FAQ are the place to confirm the current limits.
We do not offer a batch mode, an API for schools or a way to upload a folder. If you need that, see the last section.
Should you check every essay in a class set?
Think before you do. A detector has a non-zero error rate, and the more texts you run through it, the more wrong flags you will have to deal with. Which kinds of writing are most at risk is covered in AI detection false positives in student writing. Scanning a whole class also treats every student as a suspect, which changes the relationship before you have any reason to.
A narrower approach is usually fairer and cheaper in your time:
- Check pieces that already raise a question, such as a sudden change in voice, vocabulary or level compared with earlier work.
- Do not check very short submissions. Under 50 words the checker gives no result, and short texts are weak for every tool.
- Leave out lists, outlines, reference sections and templated forms, which detector vendors say are more prone to false positives. The Pangram guide lists what its maker says about this.
If your school’s policy says every submission is screened, follow the policy, and make sure it also says what a flag leads to.
How do you work through a set one text at a time?
- Prepare each text. Use the student’s own words only. Remove names and other identifying details first, because the text is sent to Pangram Labs for analysis. Our privacy page describes the data flow, and your school’s data policy decides whether student work may be sent at all.
- Prefer a file or pasted text over a scan. Scanned or image-only PDFs have no readable text and cannot be checked.
- Check one piece and read the whole result. Look at the verdict, the confidence level and the share flagged, then read the highlighted passages in context. The number alone tells you little, as what an AI detection percentage means explains.
- Write down the outcome before you open the next one. A simple table of student, date checked, verdict, which passages you read and what you did next is enough. Do not keep the student’s text in it.
- Stop at a flag, not at the end of the pile. A flagged result starts the process in what to do when an AI detector flags student writing. It does not end with a score.
What should the record look like?
Keep it factual and short. This layout works for a class set and contains no student text:
| Student (initials or ID) | Date checked | Verdict shown | Passages read in context? | Other evidence looked at | Next step |
|---|---|---|---|---|---|
| One row per student | Date | As shown | Yes / No | Drafts, earlier work | None, conversation or policy process |
The point of the record is that no student is acted on from a verdict alone. Everyone with a flag should have a “next step” that is a conversation or a review of other evidence, not a penalty.
What if you need batch scanning?
Some tools are built for it. Many schools already have a detector inside their learning platform, which scans submissions as they arrive. Which AI detector do teachers use compares the main tools on file upload and education features, and GPTZero is one of the standalone tools that describes batch scanning. Treat any figure about limits as something to confirm on the vendor’s own page, as ours are dated.
Before you adopt a batch tool, ask the questions in most reliable AI detector for teachers: what error rate it publishes by type of writing, what it does with the text, and whether it shows you which passages were flagged.
FAQ
Can I upload several files at once? No. Each check is one pasted text or one file. Upload size is limited to 10 MB per file.
Is there a bulk or API plan for schools? Not at present. We do not offer a school plan, so use the checker for individual pieces you have a reason to look at, and an institutional tool if you need to screen whole sets.
Is it safe to check student work here? Our server does not store submitted text, but the text is sent to Pangram Labs for analysis, and its own terms govern what it does with it. Check your school’s data policy first and remove identifying details. The teachers page has the full answer.
Is checking a whole class fair? Only if the policy and the follow-up process are fair. A screening result should never be used alone as the basis for a penalty. Every flag needs other evidence and a conversation with the student first.