UnTXT and GPTZero both mark AI writing sentence by sentence, and you can use either one without going through an institution first. Here is how they compare on what each actually ships today.
| Feature | UnTXT | GPTZero |
|---|---|---|
| Sentence-level detection | ✓ | ✓ |
| Named signal categories | ✓ 11 types | ✗ (probability score per sentence) |
| PDF report export | ✓ | Paid plan |
| Free tier | 1,000 words on sign-up | Limited but large free tier |
| Pricing model | Monthly plan with a CVT allowance; one-time packs for organisations | Paid plans available |
| Going over your allowance | ✓ Metered and billed, never blocked mid-document | Plan limits apply |
| Role-based interface | ✓ (Student/Teacher/Professional) | ✗ |
| Multi-currency pricing | ✓ 14 currencies | USD primary, but support available for GBP, EUR and INR |
| CVT credit gifting | ✓ | ✗ |
| Public direct access | ✓ | ✓ |
Comparison based on publicly available information about both tools, as of early 2026.
UnTXT uses eleven independently weighted detection modules. SPIKE reads AI vocabulary, SYNTAX repetitive structure, FORMAL hedging language, HUMAN the genuine markers that count in your favour, TEMPLATE boilerplate phrasing, FRONTED fronted adverbials, FORENSIC style inconsistency, RHETORIC the AI rhetorical frames, PUFF inflated and promotional language, FORMULA the challenge and future outlook template, and LEXICAL the discourse scaffolding sitting under all of it. Those eleven feed two frontier models that argue each passage out, with a third settling any disagreement, and what a reader sees is the passage marked AUTHENTIC or SUSPECT with the reasoning written out. There is no percentage anywhere in UnTXT. The full per-module breakdown is still returned to developers through the API.
GPTZero highlights at sentence level too, giving a probability score per sentence alongside one for the document. It runs its own model, trained on a corpus of AI and human writing.
The practical difference is small to describe and large in use. UnTXT tells you why a passage is flagged and what to do about it, whereas GPTZero tells you how likely it is to be AI. Neither is wrong, though they answer different questions.
UnTXT has a Teacher role with detection thresholds tuned to cut false positives on work you did not write, and it exports a report to PDF. The Educator plan carries 1,300 CVT a month, which is roughly fourteen Academic Weapon runs, and unused allowance banks to the end of your academic year instead of evaporating at the end of the month.
GPTZero has an educator product with bulk upload and classroom management. UnTXT has neither. To be clear, there is no bulk upload here and no LMS integration, so if that is what your department runs on, GPTZero is the better fit and we would say so.
UnTXT: a fixed monthly plan, from £8.40 including tax, carrying a CVT allowance. Go past it and the excess is metered at the same rate credit costs to buy outright, so an analysis is never refused halfway through a document. Unused allowance banks to your academic year or fiscal quarter rather than vanishing monthly. One-time credit packs remain, though they are for organisations of three or more. The free tier is 1,000 words on sign-up.
GPTZero: a free tier with a word limit, then monthly subscription plans at several price points.
Choose UnTXT if: you want to know why each passage is flagged rather than only how likely it is, or you want an allowance that meters instead of stopping you mid document, PDF export, or pricing in your own currency.
Consider GPTZero if: You need bulk document processing, LMS integration, or prefer a broad probability-score approach to per-signal categorisation.