The release of AI tools to the public has changed the nature of research. Now, most academics can use AI to correct and refine finished drafts. However, AI misuse is also on the rise in research, threatening ethical standards.
Text generation, data fabrication, image manipulation—these are just some of the instances of AI misuse in research. But AI tools aren’t going anywhere. Therefore, editors and scholars need to be aware of tools that help detect AI-generated content. Furthermore, authors must recognize when to disclose the use of AI in manuscript preparation.
Types of AI misuse
AI can be used to aid authors in many stages of research and manuscript preparation. This leads to various types of AI misuse, all breaching ethical standards in academic publishing. Let’s have a look at some common types of AI misuse in research.
Data fabrication
As AI advances, its capabilities grow. But with each new iteration of an AI program, more issues are created for academic publishing.
One of these issues is data fabrication. Researchers found that GPT-4 was able to produce semi-random and therefore convincing datasets in minutes. Such potentiality for data fabrication raises significant alarm bells. For example, if scientific data is used for the treatment of patients with certain health conditions, then a misleading dataset could potentially have significant consequences for treatment or rehabilitation.
AI-written abstracts
The use of generative AI is appealing to many due to its time-saving potential. In seconds, entire paragraphs can be constructed from just one or two simple prompts.
Since generative AI tools have been released to the public domain, there has been an increasing prevalence in AI-written abstracts. In fact, around one in seven biomedical research abstracts published in 2024 were believed to have been written with the aid of AI.
The misuse of AI for writing abstracts creates ethical concerns surrounding plagiarism and “hallucinations” (the outputting of false information by AI as if it were true).
Image manipulation
It’s not only text that AI can falsify. AI tools are also able to manipulate images.
There have been examples of hundreds of papers containing AI-manipulated images of Western blots, a visualization of a chemical reaction or fluorescent tags indicating the presence of specific proteins.
This type of AI misuse, like data fabrication, is misleading and threatens the integrity of scientific publishing.
AI vs. AI: How detection tools can help
Just as the misuse of AI tools raises ethical concerns, AI tools can also be used to uphold ethical standards in academic publishing.
Publishers like MDPI are adopting image integrity tools such as Proofig AI into their workflows to detect instances of image manipulation. Powered by a vast database, Proofig AI utilizes advanced algorithms to crosscheck images and identify patterns unique to AI-generated content.
For assessing the text of academic manuscripts, Gotcha GPT can distinguish between AI- and human-generated English manuscripts by using classifier models. There are also free, ready-to-use AI detection tools available online for editors and scholars to use, such as AI Text Classifier and ZeroGPT.
There are plenty of AI tools out there for editors and scholars to utilize. It’s recommended to use a few in unison and crosscheck them to contrast detections and evaluate the extent of AI usage in a paper.
Disclosing the use of AI in manuscript preparation
The misuse of AI is only going to become more prevalent within academic publishing. That’s why authors need to be aware of how to disclose the use of AI tools in manuscript preparation.
The use of generative AI must be acknowledged in a manuscript if it has been used in the process of creating the academic work. This means that the usage needs to be generative (drafting ideas or structuring material) rather than supportive (i.e., correcting spelling errors). The latter instance doesn’t need to be disclosed.
Disclosures must be made in the ‘Methods’ section of a manuscript. Reach out to a publisher or their editorial team if unsure about whether your use of AI needs to be disclosed.
Publishing your research with Preprints.org
As an author, you want to make sure you can trust whoever you submit your research to. This includes knowing they are following ethical standards when it comes to AI and authorship. Preprints.org strictly follows COPE standards concerning the use of Artificial Intelligence (AI) and AI-assisted technology in manuscript preparation.
Want to submit with us? At Preprints.org, we empower researchers to freely and instantly share their work with a global audience, helping you gain early feedback, boost visibility, and accelerate discovery. Join over 350,000 researchers advancing open science on our accessible, multidisciplinary platform. Ready to submit? Upload your preprint today and make your work quickly discoverable.
Just exploring? Browse over 100,000 preprints across disciplines and stay ahead of the latest research.
