AI text detection
Detect signs of AI writing in your text
0 words · 0 characters
AI text detection is temporarily unavailable.
The analysis service is not responding. Nothing will be charged. Please try again in a few minutes.
Paste text to identify passages with signs of AI writing in seconds.
How Veread works
How Veread reads a text
Veread splits text into tokens, reads each one in the context of the whole passage, and scores every token. Token scores are then grouped by sentence, so the result shows which passages carry signs of AI writing instead of a single number for the whole text.
Text · Veread trace
Languages
Supported languages
Analyze text in English, French, German, Spanish, Chinese, Vietnamese, and Arabic.
For students and teachers
Grammar and plagiarism checks in the same tool
Switch the check above to look for spelling, grammar and punctuation errors, or to find passages that match public web pages and Wikipedia. Both work in English, Spanish, French and Arabic and are part of the Pro plan.
Grammar check
Each suggestion is marked in the text with the correction and a short reason, by category: spelling, grammar, punctuation, word choice and style. Copy the corrected text when you are done.
Plagiarism check
The share of matching text, every matching passage with a link to its source, and quotations and the reference list set apart. Matching text is not a verdict; review each passage in context.
Accuracy
Evaluated on held-out text in seven languages.
Veread, Fileverity’s own text model, identifies AI-writing signals in seven languages. Each language is evaluated on held-out text that was not used during development.
Human writing
0.28%
of human-written texts were flagged for AI-writing signals across seven languages.
Fully AI-written
98.6%
of fully AI-written texts were detected.
AI-written and AI-assisted
95.9%
of AI-written and AI-assisted texts were detected, including fully generated, continued, rewritten and edited text.
Models excluded from development
94.9%
of texts from AI models never used in development were detected.
Held-out validation of Veread 1.1, November 2026Explore evaluation results
| Language | False positives | Fully AI-written | AI-written and AI-assisted |
|---|---|---|---|
| All seven languages | 0.3% | 98.6% | 95.9% |
| English | 0.2% | 95.1% | 94.5% |
| French | 0.5% | 99.1% | 96.8% |
| German | 0.3% | 100.0% | 94.1% |
| Spanish | 0.4% | 100.0% | 96.0% |
| Chinese | 0.0% | 100.0% | 98.8% |
| Vietnamese | 0.5% | 95.5% | 92.8% |
| Arabic | 0.1% | 98.5% | 97.5% |
Documents
Review the complete document
Bring text analysis, embedded photos, retained revisions and signature checks together in one document verification report.
To check a single image, use AI image detection.
By model
Detectors focused on model families
Five detectors use the same Veread engine, each focused on the writing patterns of a particular model family.
Console and API
Integrate text detection into your software
Analyze text in the Fileverity Console or integrate the same checks into your systems through the API.
Frequently asked questions
What does AI text detection do?
The detector assesses whether plain text shows patterns of human writing, AI writing or a mixture of both. It highlights passages with AI-writing signals so you can review them in context.
Which languages does AI text detection support?
The detector supports English, French, German, Spanish, Chinese, Vietnamese, and Arabic. Select the language or use automatic language detection.
How much text do I need?
Use about 100 words of continuous text so the model can evaluate writing patterns. Each check supports up to 100,000 characters.
What does a check cost?
AI text detection is included in every plan: a free account covers 10,000 characters a month, Personal 500,000 and Pro 2,000,000. Sign in to run a check. Beyond your monthly allowance, a check uses 2 credits per 1,000 characters or part thereof, from credit packs you can buy when needed.
What does the result indicate about how a text was written?
The result assesses AI-writing patterns and identifies passages for closer review. It does not identify the author. Before making decisions that affect people, consider the findings alongside human review, context and other evidence.
How is this different from a plagiarism checker?
The two answer different questions. AI text detection examines writing patterns and can flag original wording generated by a model; the plagiarism check finds passages that also appear on public web pages or Wikipedia and links each to its source. Both run on the same text in the same window, for English, Spanish, French and Arabic.
How is this different from the document report?
AI text detection analyzes content you paste into the tool. The document report examines an entire file: it analyzes text with Veread, checks up to 15 embedded photos and scanned pages, and reviews retained revisions, hidden content and signatures.
What does the grammar check find?
Spelling, grammar, punctuation and word-choice errors in English, Spanish, French and Arabic. Each suggestion is marked in the text with the correction and a short reason, and you can copy the corrected text. Suggestions come from a language model, so review them before accepting.
What does the plagiarism check search?
Public web pages and Wikipedia in the language of the text. The result shows the share of matching text, every matching passage with a link to its source, and sets quotations and the reference list apart. It does not see books, paywalled journals, other students’ papers or private databases, and it does not find paraphrased or translated text. Matching text is not a verdict.
How much do the grammar and plagiarism checks cost?
Both are part of the Pro plan and use credits from the plan: a grammar check 1 credit per 5,000 characters and a plagiarism check 1 credit per 1,000 characters, or part thereof. Pro adds 240 credits a month, and you can buy more at any time. Verified teachers have Pro. Checks that cannot be completed are not charged.
Guides
Practical guidance for review teams.
What each check establishes, where it stops and which evidence to request next.
Evaluating detection
Which AI Wrote This? How ChatGPT, Claude, Gemini, Grok, DeepSeek, Qwen and GLM Write
The phrases, punctuation and structure that set seven AI model families apart from human writers, with multipliers, the method and why they are not proof.
6 min read ·
Evaluating detection
AI Text Watermarks in 2026: Who Marks ChatGPT, Claude, Gemini and Grok Text
Which AI vendors watermark text in 2026, from OpenAI textGrain to SynthID, who can verify it, what the EU AI Act requires and what a mark proves.
7 min read ·
Evaluating detection
What to Do When an AI Detector Flags a Student's Essay
How teachers can review a flagged essay: read passages in context, examine the writing process, talk with the student and document a fair decision.
11 min read ·
Evaluating detection
How to Check a Whole Class Set of Essays for AI at Once
Review class submissions for AI writing in one batch. Download steps for Classroom, Canvas, Moodle, Teams and more, with guidance on interpreting results.
9 min read ·
Documents
Document Fraud Detection: AI-Generated and Altered Files
Spot fake utility bills, pay stubs and bank statements, then confirm with the issuer. Compare document verification services and see where deepfake fraud fits.
22 min read ·
Evaluating detection
How to Choose an AI Detection API: Ten Checks Before You Integrate
A buyer's checklist for AI detection APIs: input limits, sync or async results, webhooks, evidence output, data retention, test mode and pricing.
17 min read ·
Bring clear evidence to your file review.
Start with your own files, or talk to us about your team’s review process, volumes and integration needs.