Find the sentences that read as AI-written, then rewrite them as a human
If a teacher, an editor, a client or another model told you to stop writing like an AI, this is the tool for that. Paste the draft that got flagged and Detect shows which sentences read as AI-written, which look AI-assisted, and which read as human, with percentages for each.
Rewrite as Human takes the same draft and rewrites it in a human register through Grok or Rhodes. You tell it what the writing is for, such as a college essay, a cover letter or site copy, and pick a voice; the default is the Pragmatist Guide style. Detect + Rewrite runs both in one step and returns the detection result alongside the rewritten text, which you can feed back in for another pass.
The same three operations are available as an API for your own agents and pipelines: POST /api/writing-tools/detect, rewrite and humanize with your Reality Fabricator API key in the X-API-Key header. Detection runs on Pangram and is billed per 100 words; rewriting is billed at the model's normal rate plus a small tool surcharge. Source text is never logged. The tool is in alpha and requires sign-in.
Open Human Writing Tools on Reality Fabricator
What you can do
- Sentence-level detection with AI, AI-assisted and human percentages and highlights
- Rewrite through Grok or Rhodes in a human register, with the writing's purpose as context
- Selectable voice, defaulting to the Pragmatist Guide style
- Detect + Rewrite in one step, with the output reusable as new input
- API endpoints for detect, rewrite and humanize using your API key
- Source text is never logged; billing shown per operation
How it works
- Paste the flagged draft and describe what the writing is for.
- Run Detect to see which sentences read as AI, AI-assisted or human.
- Pick a voice and a rewrite model, then run Rewrite as Human or Detect + Rewrite.
- Review the rewritten text, reuse it as input for another pass, or call the same endpoints from your own code.
Frequently asked questions
How does detection work?
Detection runs on Pangram. The draft is classified and returned with window highlights, so each sentence is marked as AI-written, AI-assisted or human, along with overall percentages. Source text is not logged.
Which models do the rewriting?
Grok or Rhodes, selectable per run. The rewrite takes the purpose you give it and a chosen voice; the default voice is the Pragmatist Guide style. Output is returned as plain prose.
What does it cost?
Detection is billed per 100 words of input. Rewriting is billed at the chosen model's normal token rate plus a small tool surcharge. Detect + Rewrite charges both legs. You need to be signed in with tokens available.
Is there an API?
Yes. POST /api/writing-tools/detect, /rewrite and /humanize accept JSON with your API key in the X-API-Key header, so agents and pipelines can run the same detection and rewriting without the web page.
Does it filter content?
No. The writing tools do not apply a content filter to the text you submit. They detect and rewrite whatever draft you paste, and the source text is never written to logs.