Browser tool

Estimate AI-Likeness with Explainable Metrics

Paste an excerpt, toggle academic or strict modes, and receive a heuristic AI-likeness estimate with highlighted writing patterns and an exportable report.

Paste text and click Analyze.

Quick instructions

  1. 1

    Paste or type text

    Import text via paste, keyboard, or the sample button. Longer passages (300+ words) return more reliable signals.

  2. 2

    Choose analysis mode

    Enable academic or strict mode to adjust thresholds before running the analysis, especially for technical or highly structured writing.

  3. 3

    Run the analysis

    Click “Analyze” to generate the heuristic score, key metrics, and highlight overlays instantly on the right-hand panel.

  4. 4

    Review & export

    Copy the summary, export JSON for record keeping, or copy highlighted HTML for manual review in another application.

Examples & tips

Audit a press release

Paste the announcement copy, review highlighted boilerplate, and share the HTML export with PR leads for manual refinement.

Assess community submissions

Queue user-generated stories, run strict mode, and archive the JSON output as part of your moderation notes.

Coach a student rewrite

Highlight repetitive phrases, discuss alternative structures, and encourage mixing sentence lengths to improve burstiness.

Useful details

  • Combine the checker with the Word Counter to review readability before publishing.
  • Maintain a manual audit trail whenever AI detection influences policy or grading decisions.

About this tool

AI-generated prose is everywhere, but blunt detection claims often lack transparency. Our checker surfaces the signals behind its heuristic score so editors, educators, and compliance teams can weigh context before making decisions.

The analyser runs entirely in your browser. No text leaves your device, which is critical when reviewing embargoed briefs, student submissions, or confidential copy drafts.

Questions

Is the score definitive?
No detection method is perfect. Treat the score as a signal and combine it with human judgement and other evidence.
Why is short text unreliable?
Short passages provide limited data for statistical analysis. For best results, analyse 300+ words.
Does the tool store submissions?
Nothing is uploaded or stored. Refreshing the page clears your text and results.
Can I automate this workflow?
Use the JSON export to feed results into your moderation systems or data pipelines.

Have feedback?

Found a bug or have an idea to improve this tool?