Building with an AI agent? Copy the full API reference as Markdown, or point it at /llms.txt.

POST

Humanization API

Rewrite AI-generated text to read naturally and bypass all major AI detectors including Turnitin, GPTZero, Copyleaks, Originality.ai, and Winston AI.

Overview

The Phrasly Humanization API transforms AI-generated content into natural, human-sounding text that is indistinguishable from content written by a person. It is designed for developers and businesses who need to process AI-written text at scale while preserving the original meaning, tone, and context.

Unlike generic paraphrasing tools, the Humanization API uses purpose-built models trained specifically for AI text rewriting. The output is optimized to bypass all major AI content detectors -including Turnitin, GPTZero, Copyleaks, Originality.ai, and Winston AI -making it the most effective and affordable humanization solution for enterprise use.

The API accepts text between 20 and 5,000 words per request. Accounts with Phrasly Ultra can choose one of five rewrite modes, from gentle to max, to control how much the text changes, and turn on optional text settings such as proofreading. Streaming is supported for real-time delivery of results.

POST https://business.phrasly.ai/api/v1/humanize

Request Body

text

stringrequired

The text to humanize. Must be between 20 and 5,000 words.

mode

string

Phrasly Ultra mode, from fewest changes to strongest rewrite: gentle, light, balanced (default), strong, max. Any other value runs balanced. The standard model ignores it.

stream

boolean

Stream tokens as they generate. Default: false

model

string

Only for accounts with Phrasly Ultra enabled, which are served Ultra by default: pass latest to use the standard model instead. Ultra is available by request, and requests it serves are billed at $0.30 per 1,000 words. Request Ultra access.

more_commas

boolean

Adds the commas the model tends to leave out, so sentences read with more natural pauses. Phrasly Ultra only, +$0.01 per 1,000 words (details).

sentence_length

string

"longer" or "shorter". Nudges the average sentence longer (fewer, fuller sentences) or shorter (punchier ones). Phrasly Ultra only, +$0.01 per 1,000 words (details).

cleaner_grammar

boolean

Fewer slips in punctuation and articles. Detection bypass can dip slightly on balanced and strong. Phrasly Ultra only, +$0.01 per 1,000 words (details).

proofread

boolean

A second pass after the rewrite fixes missing apostrophes, capitals (names, weekdays, months) and missing commas, without changing a word. English text only. Phrasly Ultra only, +$0.04 per 1,000 words (details).

Response

Returns plain text (Content-Type: text/plain) with the humanized output.

Code Examples

bash

curl -X POST https://business.phrasly.ai/api/v1/humanize \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Your AI-generated text here..."
  }'

Ultra Modes

For accounts with Phrasly Ultra, mode decides how hard the text is rewritten. The standard model ignores it. Phrasly Ultra is available to Business accounts by request, at $0.30 per 1,000 words: Request access.

Mode	Description
gentle	Fewest changes. Stays closest to the original wording while still rewriting for detectors.
light	A light rewrite that keeps most of the original structure.
balanced	The default. A good mix of natural writing and detection bypass.
strong	A heavier rewrite, for text a detector still flags.
max	The strongest rewrite, for the toughest detectors. Changes the most.

Ultra Text Settings

Accounts with Phrasly Ultra can tune the writing itself, independently of the mode. Every setting is off unless you send it, and each one that runs adds its price to your Ultra rate for that request.

Setting	What it does	Add-on price
more_commastrue	Adds the commas the model tends to leave out, so sentences read with more natural pauses.	$0.01 / 1K words
sentence_length"longer" or "shorter"	Nudges the average sentence longer (fewer, fuller sentences) or shorter (punchier ones).	$0.01 / 1K words
cleaner_grammartrue	Fewer slips in punctuation and articles. Detection bypass can dip slightly on balanced and strong.	$0.01 / 1K words
proofreadtrue	A second pass after the rewrite fixes missing apostrophes, capitals (names, weekdays, months) and missing commas, without changing a word. English text only.	$0.04 / 1K words

You are only charged for a setting that ran. The standard model ignores them all: accounts without Ultra, requests with model "latest", and the rare request that falls back to the standard model pay no add-ons.
Proofread is charged only when its pass completes. Text that isn't English comes back unproofread and is not charged for it.
With stream: true and proofread on, the response still streams, but the text arrives in one piece once proofreading finishes, since it corrects the finished rewrite.

For example, at Ultra's $0.30 per 1,000 words, a request with cleaner grammar and proofread costs $0.35 per 1,000 words.

bash

curl -X POST https://business.phrasly.ai/api/v1/humanize \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Your AI-generated text here...",
    "mode": "balanced",
    "cleaner_grammar": true,
    "proofread": true
  }'

Supported AI Detectors

The Humanization API is continuously tested and optimized against every major AI content detection platform. Our purpose-built models are regularly updated to maintain consistent bypass rates as detectors evolve. Supported detectors include:

Turnitin

The most widely used academic integrity platform, trusted by over 16,000 institutions worldwide.

GPTZero

A popular AI detection tool used by educators, publishers, and content platforms to identify AI-authored text.

Copyleaks

Enterprise-grade AI content detection with multi-language support used by businesses and universities.

Originality.ai

An AI detection and plagiarism checking tool designed for content marketers, SEO agencies, and publishers.

Winston AI

An AI content detection solution with document and image scanning capabilities for teams and organizations.

ZeroGPT

A free AI text detector that analyzes writing patterns to classify content as AI-generated or human-written.

Common Use Cases

The Humanization API is used across a wide range of industries and applications where AI-generated content needs to sound natural and authentic:

Content marketing platforms that generate blog posts, product descriptions, and ad copy with AI and need output that reads naturally.
SEO agencies producing high-volume content that must pass AI detection checks before publication.
EdTech platforms and writing assistants that help students improve AI-drafted essays into polished, human-quality work.
SaaS products offering AI writing features where the final output needs to be indistinguishable from human writing.
Enterprise teams processing internal communications, reports, and documentation generated by large language models.