Undetectable AI shows up constantly in "bypass AI detection" searches, but what it actually is gets flattened to a one-line pitch. It's two features in one product — an AI-content detector and a text-modification tool — and independent testing of both has produced a more complicated picture than the marketing does. Here's what's documented, sourced, and what that means if you're looking for an alternative.
Trusted by students at
Per its Wikipedia entry, Undetectable AI is software built to identify and alter AI-generated text — so it's really two tools bundled together: an AI detector, and a separate feature that rewrites text to read less like it came from a language model. It was built by Bars Juhasz, Christian Perry, and Devan Leos, and publicly released in May 2023.
That dual nature matters, because "does Undetectable AI work" isn't one question — its detector and its rewriting feature have been tested separately by different researchers, with different results.
In July 2023, researchers at Magna Græcia University found that text run through the software was significantly harder for AI detectors to flag as AI-generated. A separate August 2023 study led by Christoph Bartneck found that common AI detectors successfully caught raw ChatGPT output, but were less successful at catching text that had been processed through Undetectable AI — broadly consistent with the Magna Græcia findings, from the same window of time.
A more recent test paints a different picture of the product's own detector specifically. In October 2025, journalist David Gewirtz retested Undetectable AI's detector feature and reported poor results — including rating genuinely human-written text as likely AI-generated, and AI-written samples as likely human. That's a test of the detector half of the product, not the rewriting half, but it's a reason to treat any current accuracy or bypass-rate claims about the product skeptically rather than at face value, regardless of which feature is being discussed.
Tools built to evade AI detection have drawn real academic scrutiny beyond just accuracy testing. A November 2023 paper by Erik Piller of Nicholls State University specifically examined the ethical implications of software like this in professional writing contexts. Separately, using any AI humanizer to submit AI-generated academic work as entirely your own is still an academic integrity issue at most schools regardless of which tool produced it or how well it evades detection — that's a policy question your institution answers, not something a rewriting tool changes.
StudyPilot's AI Humanizer is built around a different premise: making your own drafted-with-AI-assistance writing read more naturally, not promising to defeat any specific detector. It offers five rewrite modes (Fluency, Simple, Formal, Creative, and Stealth), accepts file uploads in PDF, DOC, DOCX, or TXT, and is available on a free plan, with paid Basic and Student tiers for higher usage. It doesn't market an "undetectable" guarantee, since — per the testing above — the honesty of that claim for any tool in this category is genuinely unresolved and has shifted over time even for the product that popularized the term.
Paste your text into StudyPilot's AI Humanizer and pick a mode — free to start, with no claims about defeating any specific detector, just more natural-sounding writing.
Try the Humanizer Free →