rel=nofollow citations. The promoted call-to-action above is a StealthWriter affiliate placement with its own disclosure.Does AiTextools bypass GPTZero? Before weighing the claim, check when it says it was published. Its homepage is dated three years before its own domain existed.
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
- We pulled the schema date from every URL on the site. 44 of the 51 dated pages predate the domain — registered 11 January 2026, confirmed at Verisign today.
- The earliest date, 15 January 2023, is on the homepage itself. That is three years before the website existed.
- Its GPTZero page quotes two different ranges in the same breath: scores dropping from 80–98% to under 20%, and from 70–98% to under 20%. No sample size, no date, no model, no repeat trials.
- No GPTZero screenshot exists anywhere on the site. Its only non-logo image, the social preview, returns a 404.
- It describes GPTZero as scoring on “perplexity and burstiness”. That is GPTZero’s retired 2023 framing — and the third tool this week we have found still selling against it.
- Marketing says the rewriter is “optimized to produce results that pass each of these”, naming GPTZero. Terms section 16 says “We do not guarantee that rewritten text will pass any specific AI detection system.”
- Its six before-and-after claims use six different source texts. The GPTZero one is a climate-change passage, so it cannot be compared with any of the other five.
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 AiTextools bypass GPTZero?
There is a claim, and it does not survive the first two checks anyone should run on a claim: when was this published, and what is it based on.
AiTextools is not shy about GPTZero. Across 66 pages the detector is named 1,311 times, second only to Turnitin, and it has a dedicated GPTZero humanizer page. What sits behind that page is the problem.
This page reports no first-party test of the product. Every check below was run on 6 September 2026 and is reproducible from public records.
What does the date census show?
The domain aitextools.com was registered on 11 January 2026 at 13:27 UTC. That is the registration event in Verisign’s registry record, which we queried directly today.
We then fetched all 66 URLs in the sitemap and extracted the datePublished value from each page’s structured data, programmatically rather than by sampling.
| Check | Result |
|---|---|
| Domain registration (Verisign RDAP) | 11 January 2026 |
| URLs crawled | 66 |
| Pages carrying a schema date | 51 |
| Dated before the domain existed | 44 of 51 |
| Earliest date found | 15 January 2023 |
| Page carrying that date | the homepage |
| Dates by year | 2023: 4 / 2024: 6 / 2025: 34 / 2026: 7 |
Roughly 86% of the dated pages predate the website. A site eight months old presents itself as having published continuously since January 2023, and the single oldest claim is on the front page.
What does the GPTZero page actually claim?
Two things that do not match, and no way to check either.
The page is headed “GPTZero Humanizer — Free Tool to Humanize AI Text” and promises to “rewrite flagged text to reduce GPTZero AI scores instantly.” Then it says “Most drafts fall from 80–98% AI to under 20% after one pass”, and separately “Most users see GPTZero scores drop from 70–98% AI down to under 20%.”
Two different starting ranges for the same claim, on the same page. Neither carries a sample size, a test date, a generating model, a scan mode or a repeat trial, and the page shows no GPTZero screenshot. The only evidence offered is one before-and-after pair, rendered as text, on a passage about climate change.
Can you check GPTZero yourself?
Turnitin sells no individual licences and shows students no AI report, so nobody outside an institution can produce a genuine Turnitin screenshot. GPTZero removes that excuse entirely. Anyone can paste up to 10,000 characters into gptzero.me right now with no account, no card and no signup; a free account raises the per-scan limit to 150,000 characters with a 10,000-word monthly quota, and paid tiers start at $9.99 a month.
So when a humanizer claims to beat GPTZero and shows you nothing, that is a choice, not a constraint.
It also means you can check your own text. A genuine current GPTZero result shows a three-class classification — human, AI or mixed — a confidence category rather than a bare percentage, a probability breakdown across all three classes, and sentence-by-sentence highlighting. An image showing only “X% AI”, or one displaying perplexity and burstiness, is either fabricated or dates to roughly 2023.
Which makes the absence structural rather than incidental. There is no detector screenshot anywhere on the site. The only non-logo image it hosts, its own social preview card, returns a 404.
Are the six results comparable to each other?
No, because each one uses a different passage.
| Detector page | Claimed before → after | Source text topic |
|---|---|---|
| GPTZero | 98% → 2% AI | Climate change |
| ZeroGPT | 100% → 0% AI | The future of AI in business |
| Originality.ai | 99% → 2% AI | SEO |
| Copyleaks | 97% → 3% AI | Project management |
| QuillBot | 96% → 3% AI | AI in education |
| Winston AI | 94% → 4% AI | Content marketing |
| Turnitin | no result claim at all | — |
Six topics means six different difficulty levels. Nothing here can be read against anything else, and the descending sequence from 100% down to 94% reads as a ranking when it is six unrelated single runs.
Is the technical description current?
No. The page says GPTZero scores on “perplexity and burstiness” — “how predictable your words are and how much sentence length varies” — and that AI text scores low on both.
That was GPTZero’s 2023 framing. Its current FAQ uses neither word, describing classification into HUMAN_ONLY, MIXED or AI_ONLY with confidence scores, AI-vocabulary highlighting and sentence-by-sentence highlighting, with every live result stamped with a model version.
Where do the marketing and the contract disagree?
The contact page states: “The platform checks content against multiple detectors including GPTZero, Originality.ai, Copyleaks, ZeroGPT, and Winston AI. The text rewriter is optimized to produce results that pass each of these.”
Section 16 of the Terms states: “We do not guarantee that rewritten text will pass any specific AI detection system.” Section 14 adds: “Rewritten text may still be detected by some AI detection tools in some cases. We make no guarantees about accuracy, reliability, or effectiveness of our tools.”
In fairness, the gap here is narrower than most in this series. “Optimized to” is softer than the 99% guarantees competitors print, and AiTextools publishes no headline bypass percentage at all. Its only self-performance figure, “95%+ accuracy”, describes its own detector product rather than its humanizer.
Has any detector company tested it?
None has. AiTextools is absent from the appendix table in GPTZero’s February 2026 paper, which names nine bypass services individually; absent from the nineteen tools in Pangram’s DAMAGE paper; and absent from the twenty in Pangram’s August 2025 benchmark. A full census of all 66 pages also finds Pangram mentioned zero times, including in its own Top 10 AI Detection Tools 2026 listicle, which we cover in does AiTextools bypass Pangram.
How reliable are the numbers on either side?
GPTZero’s own FAQ concedes: “Our classifier is not trained to identify AI-generated text after it has been heavily modified after generation.” Its developers page simultaneously claims a “Paraphraser Shield” means “even if AI content has been altered to look more human-like, GPTZero can detect it.” Both statements are currently published.
The two leading detector vendors publish opposite numbers on the same question. GPTZero’s February 2026 paper reports 93.5% recall on 1,000 texts run through nine bypasser services, and puts Pangram at 49.7%. Pangram’s DAMAGE paper reports GPTZero at 60.04% on humanized text, and 34.53% at GPTZero’s own default threshold. Each benchmark shows its publisher winning, and neither is peer-reviewed.
On false positives, GPTZero was one of seven detectors in Liang et al. (2023), which reported a 61.22% average false positive rate across those seven on 91 TOEFL essays. No per-detector figure was ever published, so that number is not GPTZero’s. GPTZero now claims it has cut its own TOEFL false positive rate to 1.1% — a figure no third party has verified.
datePublished value programmatically — a census, not a sample — then compared every date against the domain’s registration event as recorded in Verisign’s RDAP database. We read the GPTZero humanizer page, the six detector pages, the contact page and the Terms at source, quoting each verbatim and recording the presence or absence of a percentage, sample size, method, date and screenshot. We requested the site’s social preview image directly to confirm its status. We compared the site’s technical description of GPTZero against GPTZero’s own current FAQ language. We have no commercial relationship with AiTextools.What should you take from this?
AiTextools is more restrained than most of this market. It publishes no headline bypass percentage, its marketing language is optimized to rather than guaranteed, and its terms are candid that rewritten text may still be detected.
But its GPTZero page gives two incompatible ranges, cites one climate-change paragraph as its entire evidence, shows no screenshot for a detector that is free to screenshot, describes that detector using vocabulary it retired three years ago, and sits on a site where 44 of 51 dated pages claim to predate the domain, the oldest of them being the homepage.
Check the date stamp before you weigh any bypass number. Then check your own text: gptzero.me takes 10,000 characters free with no account, and the result it returns will be stamped with today’s model version and today’s date.
If your real worry is being wrongly flagged on your own writing, keep your drafting history — it costs nothing and it survives a detector being wrong. 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 AiTextools review, and the same question against Turnitin and Pangram. Also in this cluster: Text Polish, HueWrite, AuraWrite.
Get the free prompt pack instead
The manual rewriting prompts that lower AI signals without a subscription — and without betting a submission on a number nobody has reproduced. Free, no card.
Frequently asked questions
Does AiTextools bypass GPTZero?
Nothing checkable establishes it. The GPTZero page claims scores fall to under 20% after one pass, but publishes no sample size, no test date, no generating model, no repeat trials and no screenshot. The only before-and-after example is a single climate-change passage rendered as text. GPTZero has never tested AiTextools, and no independent party has published a result on it.
What is wrong with the dates?
They mostly precede the website. The domain aitextools.com was registered on 11 January 2026, which we confirmed against Verisign’s registry record on 6 September 2026. We then pulled the JSON-LD datePublished value from all 66 URLs in the sitemap. 51 carry a date, and 44 of those are earlier than the registration. The earliest is 15 January 2023, and it is on the homepage, roughly three years before the domain existed.
Why does that matter for a bypass claim?
Because dates are how readers and search engines judge whether a claim has held up. A page presented as published in 2023 or 2024 reads as having a track record. Here the track record is asserted by a site that did not exist for most of the period it claims to span, about a detector that has been rebuilt several times since. The current GPTZero build stamps results as Model 4.4b.
Does its own page contradict itself?
Yes, within a few lines. It states that most drafts fall from 80 to 98% AI to under 20% after one pass, and separately that most users see scores drop from 70 to 98% AI down to under 20%. Those are two different starting ranges for the same claim on the same page. Neither carries a sample size or a date.
Is its description of GPTZero accurate?
It is out of date. The page says GPTZero scores on perplexity and burstiness, described as how predictable your words are and how much sentence length varies. That was GPTZero’s original 2023 framing. Its current FAQ uses neither term, describing instead classification into HUMAN_ONLY, MIXED or AI_ONLY with confidence scores, AI-vocabulary highlighting and sentence-by-sentence highlighting. This is the third tool we have examined this week still selling against the retired description.
What do the marketing and the terms each say?
Opposite things. The contact page states that the platform checks content against multiple detectors including GPTZero, Originality.ai, Copyleaks, ZeroGPT and Winston AI, and that the rewriter is optimized to produce results that pass each of these. Section 16 of the Terms states: We do not guarantee that rewritten text will pass any specific AI detection system. Section 14 adds that rewritten text may still be detected by some AI detection tools in some cases. In fairness, optimized to is softer language than most competitors use.
Are the before-and-after figures comparable?
No, and that is a design choice rather than an accident. Six detector pages each publish a before-and-after pair, and each uses a different source passage: climate change for GPTZero, SEO for Originality, project management for Copyleaks, AI in education for QuillBot, content marketing for Winston, business for ZeroGPT. Different topics carry different detection difficulty, so no figure can be read against any other. There is no shared control text anywhere on the site.
Has any detector company tested it?
None has. AiTextools is absent from the appendix table in GPTZero’s February 2026 paper naming nine bypass services, and absent from Pangram’s DAMAGE paper and its August 2025 benchmark of twenty tools. A full census of all 66 pages also finds Pangram mentioned zero times, including in the site’s own Top 10 AI Detection Tools 2026 listicle.
