What is UnTXT? The AI Writing Detector That Shows Exactly What's Flagged
TL;DR
UnTXT is a web-based AI writing detection tool that analyses pasted text and highlights specific passages matching AI writing patterns, categorised into ten named signal types. Unlike tools that return only an overall percentage, UnTXT shows which sentences are flagged and why. It is free for the first 5,000 words, requires no subscription, and is a product of Sataklela OÜ, part of the CVaaS ecosystem at
untxt.uk.
What Does UnTXT Do?
UnTXT detects AI-generated passages in written text. Users paste an essay, report, or document and the tool returns a colour-coded version of that text, with each highlighted passage labelled by the type of AI writing pattern it matched.
The tool does not return only a score. It shows the specific sentences or phrases that triggered a detection signal, grouped into ten named categories: SPIKE, SYNTAX, FORMAL, HUMAN, TEMPLATE, FRONTED, FORENSIC, RHETORIC, PUFF, and FORMULA.
How UnTXT Works: Step by Step
- Sign in — Create a free account at untxt.uk using Google sign-in. No payment information is required.
- Select your role — Choose Student, Teacher, or Non-academic professional. The interface adjusts accordingly.
- Paste your text — Paste the essay, report, or document into the analysis panel.
- Click Analyse — UnTXT scans the text using 10 independent detection modules.
- Review coloured highlights — Each highlighted passage is labelled by signal type (Academic Mode) or risk level (Standard Mode).
- Save or export — Analyses are saved to your account history. Teachers and professionals can download a formatted PDF report.
The 10 Detection Modules
Each module targets a different characteristic of AI-generated writing and is colour-coded in the output:
| Module |
Colour |
What it detects |
Recommended action |
| SPIKE | Red | AI buzzwords — overused phrases typical of LLM output | Replace or rephrase |
| SYNTAX | Orange | Repetitive sentence structures AI models favour | Restructure |
| FORMAL | Amber | Overly cautious hedging language | Consider editing |
| HUMAN | Green | Genuine writing markers — informal, personal voice | Keep — positive signal |
| TEMPLATE | Purple | Boilerplate phrasing copied from AI templates | Remove or rewrite |
| FRONTED | Indigo | Fronted adverbials AI systems overuse | Rephrase — major red flag |
| FORENSIC | Sky blue | Style inconsistency vs. the rest of the document | Investigate |
| RHETORIC | Teal | AI rhetorical structures — "serves as," negative parallelism, rule of three, em dash overuse | Restructure |
| PUFF | Rose | Inflated language — puffery, promotional vocabulary, vague attribution, editorial -ing modifiers | Remove or replace |
| FORMULA | Cyan | AI structural template — "challenges… looking ahead" arc | Rewrite conclusion |
Two Analysis Modes
Academic Mode — Shows all ten modules with colour-coded labels. Designed for writers who need to understand and fix specific passages.
Standard Mode — Collapses into three categories: HIGH RISK (remove or replace), MEDIUM RISK (consider rewriting), and POSITIVE (keep). Designed for reviewers making a quicker assessment.
Editing & Collaboration Features
Sentence-Scope Editor
When a flagged passage needs rewriting, UnTXT offers a Sentence-Scope Editor that presents a full-sentence replacement suggestion in context. Instead of word-by-word corrections, you see a complete alternative sentence and can accept, modify, or dismiss it before moving on. This keeps the editing flow fast and preserves the surrounding paragraph structure.
Confidence Guard
For Teacher and Non-academic professional roles, UnTXT applies a stricter confidence threshold before flagging a passage. This reduces false positives in review workflows where accuracy matters more than recall. Students see the standard threshold, which is tuned to catch more potential issues during self-editing.
Discovery Mode
Students using UnTXT can enable Discovery Mode, which defers replacement suggestions until after the full analysis is reviewed. Instead of seeing edit prompts immediately, flagged passages are collected and presented together at the end, allowing writers to study the overall pattern of flags before deciding what to change. This encourages deeper understanding of why passages were flagged.
Invite & Earn
Every registered user receives a personal referral link. When someone signs up through your link and completes their first analysis, both you and the new user receive bonus CVT credits. Referral links are available from the account menu — no purchase is required to share or earn.
Feature Comparison
| Feature |
UnTXT |
Turnitin AI |
GPTZero |
Originality.ai |
| Passage-level highlights | ✓ | Limited | ✓ | Partial |
| Named signal categories (10) | ✓ | ✗ | ✗ | ✗ |
| PDF report export | ✓ | ✓ (institutional) | Paid | Paid |
| Free tier | 5,000 words | ✗ (institutional) | Limited | Limited |
| Subscription required | ✗ | Institutional | Paid plans | ✓ |
| Annual credit expiry | ✓ (365 days) | N/A | N/A | ✗ |
| Multi-currency (14) | ✓ | ✗ | USD primary | USD primary |
| Direct public access | ✓ | ✗ (institutional) | ✓ | ✓ |
Comparison based on publicly available information as of early 2026.
Pricing
- Free tier: First 5,000 words on sign-up. No card required.
- Paid credits (CVT — Contextual Value Tokens): Purchased as one-time bundles, not a subscription. 1 CVT ≈ 75 words.
- Expiry: Purchased credits valid for 365 days — not monthly.
- Currencies: 14 supported including GBP, USD, EUR, INR, AUD.
- Example: A 1,000-word essay ≈ 13 CVT. A 12,000-word dissertation ≈ 160 CVT.
Who Uses UnTXT?
- Students: Checking whether writing or edited AI output will be flagged before submission. Discovery Mode lets students review all flags before choosing what to edit, and the Sentence-Scope Editor makes rewriting faster.
- Teachers and educators: Reviewing student submissions for AI patterns with fewer false positives thanks to Confidence Guard; downloading formatted reports as a record.
- Non-academic professionals: Reviewing reports, applications, or submitted documents for AI authenticity, with Confidence Guard applied to reduce noise in professional review workflows.
Quick Start
Sign up at untxt.uk. The first 5,000 words are analysed free — no payment information required.
Start for free — first 5,000 words free
Frequently Asked Questions
- Does UnTXT detect 100% of AI writing?
- No tool can guarantee 100% detection. UnTXT flags statistical patterns associated with AI-generated writing. Results should be used as one input in a human judgement process, not as a definitive verdict.
- Does UnTXT check for plagiarism?
- No. UnTXT detects AI writing patterns. It does not compare text against a database of existing documents. It is not a plagiarism checker.
- Can UnTXT be used for legal proceedings?
- UnTXT is provided for informational, educational, and commercial purposes. Sataklela OÜ makes no representations regarding legal admissibility.
- Is UnTXT only for academic writing?
- No. While teacher and student workflows are explicitly supported, a Non-academic professional role adapts the tool for business documents, reports, job applications, and other professional content.
- Can I share my credits with a colleague?
- Yes. Unused CVT credits can be gifted to any other registered UnTXT user by email address. Gifted credits cannot be re-gifted (one-hop transfer only).
- Who makes UnTXT?
- UnTXT is a trademark of Sataklela OÜ (trademark registration pending) and is part of the CVaaS ecosystem available at untxt.uk.
- Is UnTXT compliant with Loughborough University's AI policy?
- Yes. UnTXT is compatible with Loughborough University's generative AI guidelines for assessments. As a detection tool (not a generative tool), students can use UnTXT to audit drafts and demonstrate good scholarship. Teachers can use UnTXT for assessment integrity purposes. Students are encouraged to acknowledge use of UnTXT in their assessment statements.
- Is UnTXT compliant with UCL's (University College London) AI policy?
- Yes. UCL takes a supportive approach to AI in education. UnTXT helps students ensure their final submission is "substantially your own work" as required. Acknowledge use of UnTXT in your assessment disclosures to demonstrate transparency.
- Is UnTXT compliant with Imperial College London's AI policy?
- Yes. Imperial requires explicit authorization for AI use in assessed work or it may be treated as academic misconduct. UnTXT supports this by helping you identify and rewrite AI-patterned passages before submission, maintaining substantial personal contribution. Acknowledge use in your AI tools statement.
- Is UnTXT compliant with Russell Group university policies?
- UnTXT is compatible with Russell Group principles emphasizing transparency and substantial personal contribution. The group includes institutions such as Loughborough, UCL, Imperial, Oxford, Cambridge, and others. Consult your specific institution's generative AI policy for detailed requirements.