• Changed word limit from 2,500 words to 5,000 words per process for all users.

New MultiLang Humanizer Model

• Greatly improved bypass rates
• Grammar improved 200%+
• Current supported languages: es, fr, de, it, pt, nl, en, zh, hi, ar, bn, ru, ja, ko, tr, vi, pl, uk, th, id, ms, sv, cs, ek, he

• Slightly more reliable model outputs

New Model

• Bypass rate significantly improved
  • Entirely got past GPTZero’s paraphrasing detection. Phrasly is currently the only tool that entirely passes these checks. 😊
  • 59.1% improvement across the board in detection scores compared to our previous model, confirmed by internal benchmarks. Optimized bypass for ZeroGPT, TurnItIn and QuillBot. Passes all other major checkers.
  • Confirmed to be the most undetectable model publicly available, confirmed by internal benchmarks of competitors.
• UI Changes
  • More flexibility on model selection. Our previous models may be better in terms of formal writing, where users prioritize human sounding text over being undetectable by checkers. Use the latest model for the highest bypass rate.

These features are fully live as of 5PM EST. Contact us at [email protected] if you are having any issues. Be on the lookout for more updates by end of month. - Daniel

Recent TurnItIn Update Status

• Many users have asked if Phrasly continues to bypass TurnItIn’s latest “AI Bypasser Detection” update. The answer is yes. Our proprietary technology cannot be watermarked, unlike our competitors’ solutions, eliminating any concerns about future updates. (TurnItIn update)
• To verify these claims, we tested 100 random documents run through Phrasly’s humanizer on TurnItIn. The result is 0% AI for all of them.

V6 Humanizer

• Processes text up to 3x faster than previous versions, the fastest model in the industry.
• Generates more natural, human-like text that better adapts to the original tone.
• Achieves up to 50% higher bypass rates compared to competitors in our benchmarks.
• Specifically enhances bypass capabilities for:
  • GPTZero’s updated paraphrasing detection, addressing previous issues with certain texts.
  • GPTZero’s AI editing detection.
  • Copyleaks, with improved effectiveness.

We’re committed to keeping our humanizer model the best in the industry. We look forward to your feedback!

• Increased readability and slightly tweaked bypass rates

New Model Released

• Bypassed GPTZero’s “Paraphrasing Detection” update. (Nice try, GPTZero.)
• Bypass rates for all other AI detectors have been improved.
• Our model is entirely indistinguishable from actual human writing.
• Writing should no longer contain any errors.

Reworked the “Aggressive” and “Medium” modes.

• Aggressive: Will prioritize undetectable drafts first, which may reduce quality.
• Medium: Prioritizes more readable drafts. In our testing, this version still bypasses all significant detectors.

A completely new AI Humanizer developed

• Added sentence variations, giving you more freedom to edit your humanized text.
• Fully revamped the humanizer model, giving you more high-quality outputs and higher bypass rates.
• Fixed reports of the humanizer outputs being significantly longer than the original text. The output should now be nearly identical in length.

• Developed an in-house, industry-leading AI humanizer model to ensure OpenAI’s future watermarking updates will never affect Phrasly users. (about watermarking)
• Confirmed to bypass TurnItIn and all primary AI detectors with industry-leading bypass rates.

• Improved text quality, specifically for longer texts. You should no longer face issues with the text quality degrading as your input gets larger.

• Released the AI Humanizer model to the public.