Is this an accurate AI detector?

It is an estimate, not proof. It scores transparent stylistic patterns common in LLM output (uniform sentence length, AI vocabulary, em-dash overuse). High-quality detection is fundamentally hard — treat the score as a starting point for your own judgment, never as evidence.

Will it falsely flag human writing?

Yes, it can. Formal, academic, technical, and non-native-English writing legitimately uses safe word choices and even sentence lengths — the same traits that read as 'AI'. Never use this to accuse a student or employee. False accusations cause real harm.

Can edited AI text pass as human?

Easily. Lightly edited AI text, AI told to write casually, or text run through a humanizer can score as human. Surface heuristics are simple to defeat, so a low score does not prove human authorship.

Is my text uploaded anywhere?

No. Everything runs in your browser with plain JavaScript. Nothing is sent to a server, logged, or stored.

Why is the score unreliable on short text?

Burstiness, lexical diversity, and paragraph-evenness metrics need length to mean anything. Under ~120 words the result is low-confidence; under ~40 words it's essentially a guess, so the tool suppresses a hard verdict.

AI Text Detector (Free, In-Browser)

Paste any text and see how likely it was AI-generated, scored 0-100 from 12 transparent, explainable heuristics: sentence-length burstiness, lexical diversity (MATTR), AI vocabulary clustering (delve, tapestry, leverage), em-dash density, negative parallelisms, rule-of-three triplets, transition-phrase density, and more. Every heuristic's contribution is shown so the score is never a black box. 100% client-side — your text never leaves your browser. This is a transparent smell-test, not proof: heuristic detection produces false positives on formal, academic, and non-native-English writing, so never use it for accusations or grading.

Paste any text to estimate how AI-generated it looks. Twelve transparent, explainable heuristics run live in your browser. Use it as a hint, never as proof.

Paste text to analyze0 words

Your AI-likelihood score and a full heuristic breakdown will appear here as you type.

Read this before you trust the number

This is a heuristic estimate, not proof. It cannot prove who or what wrote anything.
False positives are common on formal, technical, or non-native (ESL) writing. Never use this to accuse or grade anyone.
False negatives are easy — lightly edited or "humanized" AI text sails right past it.
The vocabulary tells age fast as models change, so old signals lose accuracy over time.
Short text is unreliable. Under ~120 words the confidence is low and the score wobbles.
Genre matters — lists, legal text, and step-by-step instructions naturally look more "AI-like."
100% client-side and private. Nothing you paste is sent to a server.

Frequently Asked Questions

Is this an accurate AI detector?: It is an estimate, not proof. It scores transparent stylistic patterns common in LLM output (uniform sentence length, AI vocabulary, em-dash overuse). High-quality detection is fundamentally hard — treat the score as a starting point for your own judgment, never as evidence.
Will it falsely flag human writing?: Yes, it can. Formal, academic, technical, and non-native-English writing legitimately uses safe word choices and even sentence lengths — the same traits that read as 'AI'. Never use this to accuse a student or employee. False accusations cause real harm.
Can edited AI text pass as human?: Easily. Lightly edited AI text, AI told to write casually, or text run through a humanizer can score as human. Surface heuristics are simple to defeat, so a low score does not prove human authorship.
Is my text uploaded anywhere?: No. Everything runs in your browser with plain JavaScript. Nothing is sent to a server, logged, or stored.
Why is the score unreliable on short text?: Burstiness, lexical diversity, and paragraph-evenness metrics need length to mean anything. Under ~120 words the result is low-confidence; under ~40 words it's essentially a guess, so the tool suppresses a hard verdict.

Related Tools

Improve My Prompt

Paste any LLM prompt and get specific, actionable fixes against 12 best-practice rules from Anthropic, OpenAI, and Google. Free, no login.

CLAUDE.md Reviewer

Paste your existing CLAUDE.md and get specific, numbered suggestions for vague rules, missing commands, repo boundaries, and token waste.

Word & Character Counter

Count words, characters, sentences, and reading time in Markdown. Free online Markdown-aware word counter.

Link Extractor

Extract all URLs and links from Markdown text. Free online Markdown link extractor.

About AI Text Detector (Free, In-Browser)

Free AI content detector. Paste any text and get an AI-likelihood score from 12 transparent heuristics — burstiness, AI vocabulary tells, em-dash density, and more. 100% in your browser, nothing uploaded. This tool runs entirely in your browser, so your content is not sent to a server. MarkdownMe provides free, fast, and reliable Markdown tools for developers, writers, and anyone who works with Markdown.