GPT Cleanup

Free ChatGPT Detector

Detect if text was copied from ChatGPT by scanning for its unique fingerprints — HTML data attributes, zero-width characters, and invisible watermarks.

100% client-side · No text is ever sent to a server · Scans for ChatGPT-specific markers

Text to Analyze
Detection Result

Scan results will appear here...

How this works: This tool scans for technical fingerprints — hidden Unicode characters, HTML attributes, and formatting markers that AI chat interfaces inject into copied text. It does not analyze writing style or use machine learning. A low score means no technical markers were found, but the text could still be AI-generated and manually cleaned. A high score means definitive technical evidence of AI tool origin.

How ChatGPT Fingerprint Detection Works

ChatGPT's web interface at chat.openai.com uses a rich-text editor that injects proprietary HTML markup when you copy text. This includes data-start and data-end attributes that track character positions, and data-message-author-role attributes that identify the AI assistant's response. These attributes are unique to ChatGPT — no other AI tool uses them.

Beyond the HTML layer, ChatGPT's text rendering engine frequently embeds zero-width spaces (U+200B), byte order marks (U+FEFF), and word joiners (U+2060) throughout the text. These invisible characters serve internal purposes like cursor management and word boundary detection, but they persist when text is copied and pasted elsewhere.

Our detector scans both layers simultaneously: the HTML clipboard data for ChatGPT-specific attributes, and the plain text for invisible Unicode markers. This dual-layer approach catches ChatGPT fingerprints whether you paste into a rich-text editor (which preserves HTML) or a plain-text field (which preserves Unicode characters).

ChatGPT-Specific Markers We Detect

HTML data-start / data-end

Critical

Character position tracking attributes injected by ChatGPT's rich-text editor. Virtually impossible to find in any non-ChatGPT source. This is the strongest single indicator of ChatGPT text.

data-message-author-role

Critical

Identifies the source as an AI assistant response within ChatGPT's conversation structure. Only present in text copied from ChatGPT's web interface.

Zero-Width Spaces

Critical

Invisible U+200B characters embedded by ChatGPT's rendering engine for cursor tracking. Found scattered through paragraphs in patterns that don't occur in human-typed text.

Byte Order Mark (mid-text)

Strong

U+FEFF characters appearing within text (not at file start) indicate AI text processing pipeline artifacts common in ChatGPT output.

Frequently Asked Questions

Common questions about detecting ChatGPT text

How do you detect ChatGPT text?
We scan for technical fingerprints unique to ChatGPT's web interface: proprietary HTML data attributes (data-start, data-end, data-message-*) and invisible Unicode characters that ChatGPT injects during copy operations. These are concrete evidence, not statistical guesses.
What makes ChatGPT fingerprints different from other AI tools?
ChatGPT has the most distinctive technical fingerprint of any AI tool. Its web interface injects proprietary HTML data attributes (data-start, data-end, data-message-author-role) that no other AI tool uses. These attributes are the single strongest indicator of AI-generated text, earning the highest severity weight in our scoring system.
Can ChatGPT text pass this detector?
If someone cleans the text with our Text Cleaner tool or manually removes the hidden markers, the technical fingerprints will be gone. Our detector finds copy-paste artifacts, not writing style. For style-based detection, complement our tool with NLP-based detectors like GPTZero or Turnitin.
Does this work with GPT-4, GPT-4o, and the ChatGPT API?
This detector works best with text copied from ChatGPT's web interface (chat.openai.com), which is where the HTML fingerprints originate. API responses typically don't contain these markers since there's no web UI involved. However, zero-width characters and formatting patterns may still be present in API output.
Is this better than GPTZero for detecting ChatGPT?
They're complementary. GPTZero uses NLP to analyze writing patterns and can detect AI text even after editing. Our tool scans for technical evidence — hidden characters and HTML attributes — which produces definitive results with zero false positives. Use both for the highest confidence.

Remove ChatGPT Fingerprints

Found ChatGPT markers? Clean them all in one click with our Text Cleaner — removes hidden characters, HTML attributes, and invisible watermarks instantly.