1. The law is already in effect—and this is serious
On February 1, 2026, Law of Ukraine No. 4742-IX “On Academic Integrity” entered into force. For the first time, it explicitly regulates the use of artificial intelligence in academic work at the legislative level. It is important to understand that the unethical use of AI-generated results is identified as a separate type of violation—alongside plagiarism, fabrication, and data falsification.
We discussed the requirements of Law No. 4742-IX and the penalties for the misuse of AI in more detail in our previous article: "The Law on Academic Integrity and Artificial Intelligence: What Has Changed for Researchers in Ukraine".
A brief overview of the consequences of violations—to understand the scope of liability.
The following consequences are possible for undergraduate and graduate students:
- a warning or a remark;
- re-doing the work or task;
- expulsion from the university;
- revocation of the diploma (if a violation is found after the defense).
For teachers and researchers, the consequences are more significant:
- a reprimand or an internal investigation;
- restrictions on scientific and educational activities;
- revocation of an academic degree or academic title;
- Being added to the Unified Register of Academic Integrity Violators is a public record that remains on file for a long time.
In other words, this is no longer just a formality. The risks—both career-related and reputational—are very real.
2. AI isn't banned, but using it "as is" is risky
The law does not prohibit the use of artificial intelligence. It requires transparency: if AI was used in the preparation of a paper, this must be disclosed. Where exactly and how—in the methodology section, in the acknowledgments, or in the appropriate field of the journal submission form—depends on the requirements of the specific publication or institution.
But if part of your text has already been written using ChatGPT, Gemini, or another tool—submitting it “as is” is truly risky. Not only because plagiarism detectors can identify it, but even more so because such text often fails to meet academic writing standards: it’s too formulaic, lacks the author’s perspective, and frequently gets bogged down in factual details.
The author’s task in this situation is to transform “raw” AI-generated text into a fully-fledged academic piece: to paraphrase, verify facts, add citations to reliable sources, incorporate their own arguments, and make the text coherent. It is this process that is referred to as the “humanization” or “rehumanization” of the text.
It is important to clearly understand two points:
- Humanizing the text does not eliminate the obligation to disclose the use of AI. If AI was involved in the writing process, this must be noted regardless of how well the final result has been edited.
- Humanization is about quality and readability, not about “fooling” the detector.
Disclaimer: Humanization tools can help make text sound more natural, but they are no substitute for the author. After using any such service, you must carefully proofread the text, standardize terminology, verify all facts and references, and add your own conclusions and reasoning. The tool is an assistant, not the author.
3. Top 4 Tools for Humanizing Text
1. HumanizerPro
HumanizerPro is a specialized tool for rephrasing and “humanizing” AI-generated content. It is designed for writers who need to make machine-generated text sound more natural and less formulaic.
How it can be useful for a researcher: when drafting individual chapters, or when adapting overly dry or repetitive AI-generated text to a more human-like style before further manual editing.
Pros:
- A relatively good paraphrase without any major distortions of the meaning.
- You can choose the tone (formal / academic / neutral).
- Supports working with longer texts, not just paragraphs.
- A user-friendly interface with no unnecessary settings.
Cons:
- Does not automatically understand the specifics of Ukrainian legislation and academic requirements.
- Does not guarantee an academic style—it often produces a "colloquial" version.
- Even after processing, the text still requires significant editing, including checking the terminology and references.
- It may inadvertently simplify the wording, which is a drawback in academic texts.
Cost: There is a free plan with a limit of 150–300 words for testing. Paid plans: Basic — about $7/month (15,000 words/month, up to 600 words per query); Pro — approximately $12 per month (50,000 words per month, up to 1,300 words per request, built-in AI detector); Advanced — higher volume for intensive use. Monthly and annual billing options are available.
2. Undetectable AI
Undetectable AI is a service that positions itself as a tool for converting AI-generated text into "human-like" text, with the ability to check the results using several detectors at once (GPTZero, Copyleaks, Turnitin, etc.).
Where it might be useful, in general: for rephrasing overly “technical” or formulaic passages; when preparing popular science materials, blog posts, and press releases—but not for final scientific articles.
Pros:
- Flexible processing modes (more formal / more conversational / balanced).
- A variety of display styles for different types of content.
- Built-in checking using multiple detectors is useful for identifying "risky" parts of the text.
Cons:
- The service’s main focus is on “circumventing plagiarism detectors,” which directly contradicts the spirit of the Academic Integrity Act.
- The quality of the output text is inconsistent: in some cases, the result is acceptable; in others, it is incomprehensible or distorted.
- For scientific articles, it is strictly prohibited to use this without completely rewriting it by hand.
- It may misrepresent terminology or alter the meaning of individual statements.
Cost: There is a free plan with a limited word count. Paid plans depend on the number of words per month: from ~$9.99/month for 10,000 words to $14.99/month for 15,000 words with monthly billing. Annual plans are cheaper—starting at $5/month (60 $/year) for 10,000 words. You can scale up to 380,000 words per month. Custom plans are available for businesses.
StealthWriter is a rewriting tool designed to make AI-generated text less detectable by automated checks. It's easy to use and requires no complicated settings.
Where it might be useful, to some extent: for editing rough drafts and short passages, which the author then revises entirely by hand to bring them up to academic standards.
Pros:
- Simple and quick to use.
- No complicated setup required—just a few simple steps from the input text to the result.
- Suitable for quick initial processing before manual editing.
Cons:
- The resulting language often comes across as "stilted"—artificial and inflexible.
- There are many formulaic phrases and clichés that immediately catch the eye of an experienced editor.
- For an academic text, doing without thorough manual editing is simply out of the question.
- It may disrupt the logical connections between sentences when paraphrasing.
Pricing: There is a free plan (up to 300 words per query). Paid plans: Basic — $20/month (20,000 words of the Ghost model, up to 400 words per query); Standard — $35/month (50,000 Ghost model words, up to 1,000 words per query); Premium — $50/month (unlimited, up to 2,000 words per query). Words do not roll over to the next month; unused credits expire.
4. WriteHuman AI
WriteHuman AI is a tool that promises to make AI-generated text sound more natural and less robotic. It is marketed as a service designed to enhance the “humanity” of text without significantly altering its content.
Where it can be useful: as an additional processing step after the initial draft, before manual editing—to reduce the number of phrases that sound distinctly “machine-generated.”
Pros:
- A simple and intuitive interface—no training required.
- Overall, it preserves the content and structure of the source text.
- It can be useful as one of the first steps in processing before manual editing.
Cons:
- Inconsistent quality: the results vary significantly depending on the type of text.
- There may be some strange or unnatural phrasing that needs to be corrected.
- It does not take into account the specific characteristics of scholarly writing and academic terminology.
- Not suitable as the sole editing step—manual proofreading is required.
Pricing: There is a free plan (3 requests per month, up to 200 words each). Paid plans: Basic — $12/month (80 requests, up to 600 words each); Pro — $18/month (200 queries, up to 1,200 words each, priority support); Ultra — $48/month (unlimited queries, up to 3,000 words each). A discount of ~33% is available for annual payment.
4. Comparative Table of Tools
Below — compiled table for fast comparison. Home conclusion, which from her follows: all four tools auxiliary. None not provides “ready to submission» scientific text. After everyrequired chairperson author and manual editing.
5. When You Need a Real Person: Services from the “Article Academy”
Humanization tools can “smooth out” the text and remove the most obvious signs of machine generation. But they do not address several key challenges:
- do not adapt the text to the requirements of a specific journal or the Ministry of Education and Science;
- are unfamiliar with the logic of the new law and will not be able to help you properly report your use of AI;
- do not verify compliance with the requirements of Scopus, Web of Science, or Category B scientific journals;
- They are no substitute for a scholarly editor who assesses the structure, reasoning, and overall style.
This is exactly where we need live editors who understand the academic context.
"Academy of Articles" is a team of professional editors and academic consultants who work with articles in Ukrainian and English. Among the services that may be useful to authors who have already used AI in preparing their texts are:
- Editing texts partially written using AI—checking structure, logic, reasoning, and style; bringing them up to academic standards.
- Guidance on properly disclosing the use of AI—where and how to include this information in the methodology section, acknowledgments, or a cover letter to the journal.
- Adapting articles for specific journals—structure, formatting requirements, and a reference list in accordance with the journal’s style guidelines.
If you’re not sure whether your text meets both the requirements of the law and the journal’s standards, it’s better to seek professional editing than to risk your reputation. Early quality control is always less expensive than dealing with the consequences after submission.
Conclusions
The law is already in effect—and while this is no reason to panic, it is a reason to be vigilant. Starting February 1, 2026, the unethical use of AI in academic work will constitute a separate type of violation, equivalent to plagiarism. For undergraduate and graduate students, this could result in having to rewrite their work or being expelled. For researchers and faculty members, it could lead to a reprimand, restrictions on their activities, or even the revocation of their academic degree. The registry of violators is public. This means that the consequences are long-lasting and can significantly impact one’s career.
If the text is partially generated by AI, you’ll need to address this in two ways. First, declare the use of AI where required by the journal’s, university’s, or Ministry of Education and Science’s guidelines: in the methodology section, in the acknowledgments, or in the cover letter. Second, you need to “humanize” the text—that is, transform it from a machine-generated template into genuine academic material: with its own line of reasoning, verified sources, precise terminology, and the author’s original perspective. One without the other does not solve the problem.
Humanization tools are a useful intermediate step, but not a final solution. HumanizerPro, Undetectable AI, StealthWriter, and WriteHuman AI—all of them can help make text sound more natural during the initial processing stage. But none of them knows your journal’s requirements, verifies facts, spots logical flaws in your argument, or ensures the quality of your scientific conclusions. After using any such tool, you must proofread the text yourself or have it reviewed by a professional editor. Without this step, humanization tools create only a false sense of security—and this is perhaps the biggest trap authors fall into.
The best approach is a combined one: transparent and honest disclosure of AI use, humanization support services as the first step in the processing, and professional scientific editing as the final stage before submission. It is precisely this process that allows you to remain within the bounds of the law while submitting a paper that meets the standards of a specific journal or institution. Trying to cut corners at any of these stages usually ends up costing more than it seems at first.
And finally—reputation. A researcher builds it over the years through publications, reviews, citations, and standing among peers. A single violation of academic integrity, documented and entered into a public registry, can undo much of that. No AI tool and no amount of time saved is worth that much. It’s better to spend time on thorough editing now than to fix the consequences after the paper has already been submitted or defended.
