Does UnAIMyText bypass Pangram? Nobody knows — and that is the strange part, because this is the one vendor in the category that publishes its own testing method in full.
The only tool that bypassed Pangram and Turnitin in 2026.
Most humanizers clear one detector and get caught by the other. StealthWriter is the one that gets past both — run Ghost 5.2 Pro at level 7–8, section by section, and re-check before you submit.
Try StealthWriter →Key Takeaways
- UnAIMyText is the most transparent vendor in this category. It publishes sample sizes, the models behind its corpus, its exact humanizer settings, and a month-by-month results history.
- Across a full census of all 41 pages on the site, the word Pangram appears zero times.
- It has published six named detector reviews — and reviews two humanizers Pangram itself benchmarked. It has never written about Pangram.
- It claims 97% past GPTZero and 98% past Turnitin. Two competitor tests both found 100% GPTZero detection. Both tested a version of the product that no longer exists.
- Its self-run benchmark is built from 2024-era models — GPT-4o, Claude 3.5 Haiku, Gemini 1.5 Flash — run in July 2026 against 2026 detectors.
- Zero detector screenshots exist anywhere on the site. The only image is the logo. The competitor review that contradicts its numbers publishes about twenty.
- The privacy policy says it uses no analytics cookies. The homepage loads Google Analytics.
The only tool that bypassed Pangram and Turnitin in 2026.
Most humanizers clear one detector and get caught by the other. StealthWriter is the one that gets past both — run Ghost 5.2 Pro at level 7–8, section by section, and re-check before you submit.
Try StealthWriter →Does UnAIMyText bypass Pangram?
Unknown. But the reason it is unknown is different here, and worth spelling out, because the usual story does not fit.
Most tools in this cluster shout a bypass percentage in their marketing and quietly disclaim it in their terms. UnAIMyText does not. It prints the caution itself, on the homepage, directly under its own results table: “no humanizer can promise a specific score on every detector, every time, and you should be sceptical of any tool that claims otherwise.” Then it publishes a benchmark with a disclosed method and a month-by-month history.
And Pangram is in none of it. We crawled the sitemap and read all 41 URLs. The word does not occur once.
This page reports no first-party test. Everything below was fetched directly on 28 August 2026.
How much does it actually publish?
More than most detector vendors do about themselves. Its detector-tests page discloses 50 samples per round; the models that generated the corpus, broken out by count; the content-type split across academic, business, blog and creative; the exact humanizer configuration used; a 250 to 600 word range; that all detectors were run inside a four-hour window “to minimize model drift”; that error samples were discarded and replaced; and that samples are not reused between rounds. It adds a five-point limitations section and a history table.
| Detector | Raw AI detected | After Ultra | Claimed pass rate |
|---|---|---|---|
| GPTZero | 96% | 3% | 97% |
| Turnitin | 94% | 2% | 98% |
| Copyleaks | 90% | 4% | 96% |
| Pangram | never run | never run | never claimed |
Set against that, four things deserve stating plainly. The corpus is generated by 2024-era models — GPT-4o, GPT-4o mini, Claude 3.5 Haiku and Gemini 1.5 Flash — and run in July 2026 against detectors calibrated on far newer output. Only two months of the “monthly” history exist, June and July 2026, and the August row still read pending on 28 August. The prose describes testing in Standard mode while every headline figure is Ultra. And there is no raw data of any kind.
Where is the proof?
There isn’t any, in the evidential sense. We extracted every image reference across the homepage and both test pages. The only image file is the logo. Every pass rate on the site is rendered as HTML text in a table.
Here is the awkward symmetry. The vendor claiming 97% and 98% publishes zero screenshots. The competitor review that says the tool fails publishes about twenty. Neither is disinterested, but only one of them shows its working.
What did the independent tests find?
Two exist. Both are competitors, and neither tested Pangram.
A December 2025 review reported: “GPTZero flagged every single humanized output at 100% AI detection.” It disclosed three samples of roughly 200 words, all three modes, a stated re-test protocol, about twenty embedded screenshots and a link to raw data — and it closes by steering readers to the reviewer’s own competing product. A November 2025 review reported 100% AI on GPTZero, ZeroGPT, Turnitin and Copyleaks with no methodology whatsoever, and listed its own humanizer among the detectors doing the judging. A third site then recycled the second as “independent 2026 testing”.
Has Pangram named it?
No. UnAIMyText is absent from the DAMAGE paper‘s 19 named tools and from the August 2025 benchmark‘s 20. The Pangram 4 report from July 2026 anonymises its 13 commercial humanizers as “Commercial A” to “Commercial M”, chosen by top web traffic in early 2026, so presence there can be neither confirmed nor ruled out. Never named is the accurate phrasing; never tested is not.
The pointed detail is that UnAIMyText’s own blog reviews TwainGPT and JustDone — two humanizers Pangram scored by name at 92.7% and 93.5%. It has read that far into the landscape and still never written the word.
What about the paperwork?
The terms disclaim the outcome in the ordinary way: “We do not guarantee that humanized content will be undetectable by all AI detection tools at all times.” Unusually, the same sentiment appears in the marketing, so there is no gap to expose there. None of the four legal pages carries a last-updated date.
The real inconsistency is in the privacy policy. It states: “We do not use tracking cookies, analytics cookies, or advertising cookies.” The homepage HTML loads Google Analytics. It also promises text is “discarded immediately” while selling a paid tier whose feature is Save History — a contradiction the homepage FAQ quietly concedes.
Refunds run 7 days and only under 2,000 words used, which is 6.7% of a month on the 30,000-word Basic plan. The operator is “CoreAI LLC” with no state, no registration number and no address; the contact page lists “Location: Global Support” and “Call us at: Coming Soon.” The affiliate page pays 35% of the first invoice only, names no tracking platform and states no cookie window. Its subreddit has 15 subscribers and one post, by the vendor. And one homepage testimonial cites a Reddit permalink that, when fetched in full, contains no such comment.
What should you take from this?
UnAIMyText has professionalised the presentation of evidence while the evidence itself stays entirely self-generated, unaudited, and untested against the detector that matters most. That is a genuinely better posture than the tools that simply lie, and it is still not proof of anything.
What the record supports is Pangram’s own published ceiling: across 13 unnamed commercial humanizers it reports catching humanized text 97.69% of the time, with the best evader still caught 92.78% of the time. Nothing about a well-documented self-test moves a tool outside that band.
If the real worry is being wrongly flagged on your own writing, keep your drafting history — it costs nothing and beats any rewriting tool as a defence, and it matters most for the writers detectors treat worst: ESL writers and AI detection and false positives for neurodivergent students. Check before you submit with the best pre-submission check and the Turnitin self-check, and strip the obvious tells by hand first via what to remove before using an AI humanizer.
For the product itself see our UnAIMyText review. For the Pangram cells with real evidence, see Undetectable AI, StealthGPT, QuillBot, StealthWriter, Phrasly, Humbot, Rephrasy, Ryne AI, Clarity Bubble and SuperHumanizer, plus the Pangram review.
The one tool that got past both
Pangram and Turnitin catch most humanizers — one or the other, and usually the one you did not think to test. StealthWriter is the only tool that bypassed both in 2026. Run Ghost 5.2 Pro at level 7–8, section by section, and re-check before you submit.
Free plan to start, no card. Or get the free prompt pack first →
Affiliate link — I earn a commission if you upgrade.
Frequently asked questions
Does UnAIMyText bypass Pangram?
Unknown, and unusually so — because this is a vendor that publishes more testing detail than most detector companies do. It benchmarks itself monthly against GPTZero, Turnitin and Copyleaks and publishes the method. It has never run Pangram, never mentioned Pangram, and no third party has tested the pair either.
How much methodology does it actually publish?
More than almost anyone in this market. Its detector-tests page discloses 50 samples per round, the exact models that generated the corpus, the content-type split, its precise humanizer settings, a 250 to 600 word range, that all detectors were run within a four-hour window, and a limitations section. It also publishes a month-by-month history rather than a single flattering snapshot.
So why treat the numbers with caution?
Because the process is disclosed and the evidence is not. There are no screenshots, no raw samples and no per-sample data — every figure is a self-reported HTML table. The corpus is built from 2024-generation models run in July 2026. Only two months of the “monthly” history exist, and the August row still read “pending” on 28 August. And the prose describes Standard mode while every headline number is Ultra.
What did independent testers find?
Two tests exist and both are competitors selling rival humanizers. A December 2025 review found “GPTZero flagged every single humanized output at 100% AI detection” across three samples with about twenty screenshots. A November 2025 review reported 100% AI on GPTZero, ZeroGPT, Turnitin and Copyleaks with no methodology at all — and listed its own product as one of the detectors. Neither tested Pangram, and both tested a product version that no longer exists.
Is UnAIMyText in Pangram’s published research?
It is not named in any of it. Absent from the DAMAGE paper’s 19 tools, absent from the August 2025 benchmark’s 20 tools. The July 2026 Pangram 4 report tests 13 humanizers anonymised as “Commercial A” to “Commercial M”, chosen by top web traffic, so its presence there cannot be confirmed or excluded.
Does it warn users about bypass claims?
Yes, and this is where it breaks the pattern of its competitors. Its homepage says directly: no humanizer can promise a specific score on every detector, every time, and you should be sceptical of any tool that claims otherwise. That is the sceptic’s warning printed in the vendor’s own marketing, not buried in the terms.
What does it cost and can I get a refund?
A visitor gets 450 words per input free with no account; a free account gets 500. Basic is $9.99 a month for 1,000 words per input and 30,000 a month. Pro is $14.99, anchored against a struck-through $24.99, for 2,000 per input and 100,000 a month. Refunds are limited to 7 days and only if fewer than 2,000 words have been used — 6.7% of one month on Basic, 2% on Pro.
Who operates it?
“CoreAI LLC”, named consistently across the legal pages, with no state of incorporation, no registration number and no postal address. The contact page lists “Location: Global Support” and “Call us at: Coming Soon.” No corporate registry record was findable.
