AI Gen Text Detector

Paste the text, get back a report you can act on: an overall verdict, the share that scores as AI, and the flagged passages quoted with their score and confidence. Runs on Pangram's detection API, covering English and 20+ languages.

What it does

• One verdict — AI, Human, or Mixed — plus the AI / AI-assisted / human split.
• Pinpoints which passages score as AI, quoted verbatim with a score and confidence rating.
• Flags text run through a "humanizer" tool to evade detection, so a clean score doesn't fool you.
• Reads Spanish, French, Arabic, Chinese, Japanese, Ukrainian and more — quoted in the original, never translated.
• Says when a result is weak evidence instead of guessing.

What to send

The text, pasted in full, up to 3,000 words. Accuracy climbs sharply above ~150 words. Ask for a shareable highlighted-view link.

"Check this guest post before we run it."

What you get back

# AI Detection Report

**Verdict: AI Detected** — a mix of AI and human-written content.

AI-generated 34% · AI-assisted 0% · Human-written 66%

> ⚠️ Humanizer signal — 1 of 2 passages looks laundered.

Then the per-passage table and flagged excerpts.

Good for

• Vetting freelance submissions and guest posts before publishing
• Spot-checking a contributor whose voice suddenly changed
• Screening non-English copy your team can't eyeball
• Giving an editorial call something quotable to stand on

Limits

Pasted text only, 3,000-word max — no URLs, files, rewriting, or plagiarism check. Detection is a statistical signal about text, never proof about a person — the report says so. Short passages are unreliable — it says that too. If the service can't run, you get a stated failure, not a quiet "no AI found".