Decopy AI pairs an AI text humanizer with an AI content detector and a set of study tools — homework help, document and video summaries, transcription — inside one browser workspace aimed mainly at students and academic writers.
Used this tool? Rate it
Used this tool? Rate it
Decopy AI is a browser-based writing workspace that combines two functions most products keep apart: an AI text rewriter that reshapes machine-written prose so it reads as though a person wrote it, and an AI content detector that estimates whether a passage was machine-generated in the first place. The homepage presents both as co-equal entry points under the banner "Your All-in-One Writing Solution," and the navigation bar leads with "Detect Now." Around those two pillars sits a wider set of study utilities — a homework helper, document and video summarizers, speech-to-text and text-to-speech, an OCR reader, a story generator and flashcard tools — all reachable from the same account.
It is worth being direct about what the rewriting side of this product is for, because the marketing is direct about it. The AI Humanizer page states that its rewriting can bypass AI detection from ZeroGPT, Copyleaks, and Originality.ai, describing the mechanism as advanced rewriting to beat high-level AI audits. That is a claim about concealment, not about writing quality. A tool that promised to make prose clearer would be measured against readability; a tool that promises to defeat named detectors is measured against those detectors. Understanding Decopy AI means holding both descriptions at once: it is a genuine multi-tool study workspace, and its headline feature is an evasion feature.
The audience follows from that positioning. Rewrite presets include Academic, Essay, Report and Cover Letter; the homework helper covers school and university subjects; the summarizers target lecture material and long PDFs. Students and academic writers are the centre of gravity, with marketing and SEO writers a secondary group. That audience is also precisely the group for whom the concealment use case carries the most institutional risk, and this page treats that tension as a fact about the product rather than as a verdict on anyone who uses it.
Pricing is published openly, and it is worth stating plainly because several third-party write-ups describe this product as entirely free with no paid tier — a description the official pricing page contradicts directly. There are four levels. Free costs nothing and grants 10 monthly quotas with 10 premium credits, Lite is 9.99 US dollars per month for 1,100 quotas and 800 credits, Basic is 14.99 US dollars per month for 3,000 quotas and 2,000 credits, and Premium is 29.99 US dollars per month for unlimited basic usage plus 5,000 credits. Two meters run in parallel: general usage quotas, and premium credits that the higher-quality modes draw down. Paid tiers can buy additional premium credits that do not expire, and the effective credit price falls as the tier rises, from 0.0125 dollars per credit on Lite to 0.006 on Premium. Premium also advertises unlimited storage as a forthcoming addition.
The refund terms deserve attention before you subscribe, because they are more conditional than the marketing suggests. Refunds are assessed case by case against your account's actual usage records, credit packs are non-refundable once any credit has been spent, and requesting a refund does not by itself cancel the subscription. Stated grounds include accidental charges, forgotten cancellations and dissatisfaction with the service, with support replying in roughly 48 hours and approved refunds landing in about five to ten business days. The practical implication is that you should cancel and request a refund as two separate deliberate actions, and that unused credits are the only ones you can realistically expect back.
The most important limitation is not a feature gap but a direct challenge to the product's core promise. On 27 August 2025 Turnitin announced that its AI writing detection now includes AI bypasser detection, built specifically to surface text that has been deliberately altered by AI humanizer tools. Because Turnitin is the system that most universities actually run submissions through, a rewriting engine's ability to satisfy ZeroGPT or Copyleaks does not establish that it satisfies the detector that decides academic outcomes. Turnitin also noted that the bypasser capability was trained and tested for English only, and that it ships within its Originality add-on rather than in every configuration — so the exact exposure depends on what an institution licenses.
The framing from that announcement is unusually pointed, and readers weighing this category deserve to see it. Turnitin's chief product officer Annie Chechitelli framed the category bluntly, saying that these companies exist to profit from students' misuse of AI by providing free and easy access to humanizers to conceal AI-generated content. That is one vendor's characterization of a competitive landscape and should be read as such, but it comes from the organization whose detection results carry institutional weight, and it describes the business model of the category accurately enough that it cannot simply be dismissed as rhetoric.
Independent user evidence is thin and what exists is unflattering, though the sample is too small to be conclusive. Its Trustpilot TrustScore sits at 2.3 out of 5, but that figure rests on only eight reviews, far too small a sample to support any statistical conclusion about quality. What is more informative than the number is the recurring theme: reviewers report that text processed by the humanizer was subsequently rated as fully AI-written by Copyleaks, one of the very detectors the product names. The page also notes the company has not replied to negative reviews. Beyond Trustpilot, the product has accumulated almost no independent evaluation — a single review on a major directory, no Product Hunt launch page and no Capterra listing found — so anyone expecting a settled third-party consensus will not find one. Search results for reviews of this product are dominated by blogs owned by competing humanizer services, which have an obvious interest in the conclusions they publish and which this page does not treat as evidence.
There is also an unresolved tension inside the product itself. The same site tells educators that no AI detection tool can be 100% accurate, that human-written text may sometimes be flagged as AI-generated, and that a score should act as a review signal rather than the only basis for judgment. That is careful, defensible guidance about the limits of classifiers. It sits alongside a rewriting feature sold on its ability to beat named detectors. Both positions can be argued individually; together they describe a product that asks detection to be treated sceptically while selling a way to manipulate it.
On the legal and privacy side, what the documents omit matters as much as what they contain. The privacy policy names ZINGKODE LTD. as the operating entity and commits to holding personal information for no more than three months after an account is terminated, and it grants the standard GDPR and CCPA rights with a one-month response window. Nowhere in that document is there any statement about whether the text you paste in is used to train models, which is the single most consequential privacy question for a rewriting tool. The terms are similarly selective: reaching them requires knowing they live at the /terms-of-use path, since the conventional /terms and /terms-of-service addresses both return 404. The terms require users to be at least 18 years old and prohibit resale and scraping, yet they say nothing whatsoever about academic use, ghostwriting, or submitting rewritten work as your own. For a product whose headline feature is defeating academic detectors, that silence is conspicuous — and, as noted above, an absent restriction is not an authorization.
Finally, the ordinary caveats apply. Rewriting engines alter emphasis and can blur technical precision, so every output needs a human comparison pass against the original. The detector publishes no accuracy figure, so its scores should not be treated as authoritative. And the academic-integrity question is not settled by the tool's own FAQ, which argues that using AI as a learning aid is not cheating — a position about the homework feature that does not extend to concealing authorship of submitted work. Whether any particular use is permitted is determined by your institution's policy, not by the vendor's view of it.
Partly. Basic AI detection runs without an account, and the Free plan gives 10 monthly quotas and 10 premium credits at no cost. But the claim circulating in several third-party reviews that the product is entirely free with no paid plans is contradicted by the official pricing page, which lists Lite at 9.99, Basic at 14.99 and Premium at 29.99 US dollars per month. The higher-quality rewriting modes consume premium credits, so sustained use is a paid proposition.
You should not assume so. Turnitin added AI bypasser detection on 27 August 2025 specifically to identify text modified by humanizer tools, and it is the system most institutions actually use. Marketing claims about defeating ZeroGPT, Copyleaks or Originality.ai say nothing about Turnitin, and user reports on Trustpilot describe rewritten text still being scored as fully AI-generated by Copyleaks. Treating any bypass claim as reliable is a bad bet with your academic record as the stake.
That depends entirely on your institution, and the answer is not the vendor's to give. Many institutions now address rewriting tools explicitly in their academic-integrity policies, and the framing from Turnitin — the organization whose detection results most institutions rely on — treats concealment of AI authorship as misconduct rather than as a grey area. The tool's own terms of use are silent on the subject, and silence in a vendor's terms is not permission. Check your course and institutional policy before rewriting anything you intend to submit.
The humanizer accepts up to 50,000 characters per document and the detector up to 200,000 characters of input. Those are per-pass ceilings, not monthly allowances — how many passes you get is governed separately by your plan's quota and premium credit balance.
No. Searches of both Apple App Store endpoints return no listing for this product, so it is a web-only workspace accessed through a browser on any device. A browser extension exists alongside the main site, but there is no native iOS or Android application.
The privacy policy does not say. It covers data such as IP addresses, browser type and timestamps, sets a retention limit of three months after account termination, and provides GDPR and CCPA rights — but it makes no statement either way about whether submitted text feeds model training. For a rewriting tool this is a meaningful gap, and the honest answer is that the site does not address it. If your text is confidential, unpublished or subject to a third party's rights, that silence is a reason for caution.
The site publishes no accuracy figure and explicitly says no detector can be perfect, warning that human writing is sometimes flagged as machine-generated. That candour is appropriate — no consumer detector's accuracy claims should be taken at face value — but it also means you have no published basis for calibrating the score. Use it as one input among several, never as proof of authorship in either direction.
The privacy policy identifies ZINGKODE LTD. as the operating entity, without publishing a physical address. The company's Trustpilot profile, claimed in August 2025, lists contact information in Hong Kong. Corporate details beyond this — founding date, team size, funding — are not disclosed on the site and could not be verified from independent registry sources.
First, do not panic-rewrite: running original work through a humanizer to lower a score creates exactly the modification pattern that bypasser detection looks for, converting a false positive into something much harder to explain. Preserve your drafts, revision history and research notes, since contemporaneous evidence of your process is the strongest defence available. Then raise it with your instructor directly. The site's own guidance supports this approach, advising that detection results be weighed alongside drafts, writing history and the student's explanation rather than treated as decisive.
Possibly, but the terms are conditional. Refunds are decided case by case against your account's actual usage records, and credit packs cannot be refunded once any credit has been used. Critically, asking for a refund does not cancel your subscription — you must cancel separately, or you may be billed again while your request is being reviewed. Support responds in roughly 48 hours and approved refunds take about five to ten business days to appear.
