How Does UnTXT Work?

UnTXT analyses any text for AI writing patterns and returns a colour-coded annotated version, with each flagged passage labelled by the type of pattern detected. Here is the full process from sign-in to export.

Step-by-Step Process

  1. Sign in at untxt.uk Use a Google account - no separate registration. New accounts receive a free allocation covering the first 5,000 words of analysis. No payment information required.
  2. Select your role On first sign-in, choose Student, Teacher, or Non-academic professional. The interface, default analysis mode, and available features adjust accordingly. Role can be changed at any time.
  3. Paste your text Paste the essay, report, dissertation chapter, or document into the main analysis panel. Text is entered directly - there is no file upload.
  4. Run the analysis Click Analyse. UnTXT processes the text using its eleven-module detection engine. Each module operates independently.
  5. Review the colour-coded results The analysed text is returned with highlights. In Academic Mode, each passage is labelled by module type (SPIKE, SYNTAX, etc.) in its colour. In Standard Mode, highlights collapse into HIGH RISK (remove or replace), MEDIUM RISK (consider rewriting), or POSITIVE (keep - genuine writing marker). An overall risk score for the document is shown. For any flagged passage, the Sentence-Scope Editor offers a full-sentence replacement suggestion in context - you can accept it, modify it, or dismiss it and move on. Students in Standard Mode can defer all suggestions by clicking "Find Replacements" on-demand rather than auto-fetching suggestions immediately.
  6. Save or export Every analysis is saved automatically to account history (date, word count, score, credit cost). Teacher and Professional users can download a formatted A4 PDF report from the analysis view or from history.
  7. Optional - Academic Weapon ⚔ After your standard analysis, click ⚔ Academic Weapon in the toolbar. Three frontier AI models - Claude Opus 4.7, GPT-5.5, and Gemini 3.1 Pro - conduct a structured adversarial debate about your text, typically taking 2–4 minutes. Results are applied as highlights (AUTHENTIC in gold, SUSPECT in sky blue) with a full per-model reasoning panel. The highest-confidence authorship signal in the system.

The 11 Detection Modules + Academic Weapon

Each module targets a distinct characteristic of AI-generated writing:

SPIKE

AI buzzwords - overused phrasing typical of LLM output. Action: replace or rephrase.

SYNTAX

Repetitive sentence structures AI models favour. Action: restructure.

FORMAL

Overly cautious hedging language. Action: consider editing.

HUMAN

Genuine writing markers - informal expressions, personal voice. Action: keep these.

TEMPLATE

Boilerplate phrasing copied from AI templates. Action: remove or rewrite.

FRONTED

Fronted adverbials AI systems overuse. Major red flag. Action: rephrase.

FORENSIC

Style inconsistency vs. the rest of the document. Action: investigate.

RHETORIC

AI rhetorical structures - "serves as," "not just X but Y," rule-of-three triplets, em dash overuse. Action: restructure.

PUFF

Inflated language - significance puffery, promotional vocabulary, vague attribution ("experts argue"), editorial -ing modifiers. Action: remove or replace.

FORMULA

AI structural template - the rigid "challenges… looking ahead" arc typical of AI essays. Action: rewrite the conclusion with a genuine argument.

LEXICAL

AI-characteristic discourse scaffolding - discourse marker density, sentence-length uniformity, type-token ratio deviation. Action: vary sentence length and replace transitional scaffold phrases.

ADVERSARIAL ⚔

Three-model consensus via Academic Weapon. Highlights appear only when Claude Opus 4.7, GPT-5.5, and Gemini 3.1 Pro all independently agree. The highest-confidence signal in the system. Action: review per-model reasoning in the side panel.

How Credits Work

Analysis uses CVT (Contextual Value Tokens). One CVT covers approximately 75 words. Credits are not a monthly subscription - they are purchased once as a bundle and remain valid for 365 days. The free tier covers the first 5,000 words at no cost.

Invite & Earn

Every registered user has a personal referral link, available from the account menu. When someone signs up through your link and completes their first analysis, both you and the new user receive bonus CVT credits. No purchase is required to share or to earn - the referral link works with the free tier.

Confidence Guard for Reviewers

Teacher and Non-academic professional roles use Confidence Guard, which applies a stricter detection threshold before flagging a passage. This means fewer false positives in review workflows, so highlighted passages carry higher confidence. Students see the standard threshold, tuned for broader self-editing coverage.

Try it free - first 5,000 words