Current Trends of Artificial Intelligence in Academic Publishing

Publishing scholarly literature is a complex web of research, writing, submissions, and peer review. Countless hours are spent manually combing through articles, checking for language proficiency, content errors, journal compliance, and a host of other common issues. Naturally, wherever grueling effort is spent, human ingenuity finds a way to simplify, streamline, and automate. In recent years, advanced artificial intelligence (AI) techniques have emerged to assist both researchers and publishers in optimizing their workflows. Here, we will discuss some of the most recent advancements in AI technology as applied to scholarly publishing.

What is AI in scholarly publishing?

There are several goals that developers intend to reach by using AI in scholarly publishing:


Generally, AI methods use millions of scientific research papers as training data to determine patterns and assist with tasks such as language editing and technical checks. Humans perform these tasks at a much slower rate than a machine, so developing tools in this area is valuable. Although the time-consuming problems in publishing cannot be completely solved with AI, using such methods can drastically reduce the time spent on repetitive tasks and open the door to new, more productive systems.

Recent tools in the industry

A large barrier to publishing a manuscript is the quality of the language. Improving the writing before submission can reduce the time spent during review and increase the chance of publication. Thus, automated language editing software is useful for even native speakers. While basic language tools exist within applications such as Microsoft Word and Overleaf, they often fail to suggest corrections within a scientific context. New AI tools such as Writefull are using deep learning models to detect missing citations, styles issues, and science-related language problems. Further applications for AI methods like Writefull include referring authors to human language editing services (e.g., LetPub) and assigning language quality scores before and after manual editing to evaluate the proficiency of copyeditors.

Clarivate Analytics, a leader in the industry, has focused on the problems faced by academic journal editors and aimed to integrate AI into the editorial workflow. Clarivate recently developed an AI tool that crawls individual manuscripts and pre-populates forms for journal submission. The intention is to minimize the time spent filling out tedious forms for each publication; this tool also has applications in other areas such as job searches. Furthermore, UNSILO Technical Checks, an API solution integrated into manuscript management software, has made recent strides in automatically ensuring basic journal guidelines are met (e.g., whether a funding statement is included).

While language-based AI focuses on the pre-submission and submission stages of the editorial workflow, other tools are being developed for the later stages. For example, Clarivate designed Reviewer Connect, which allows journal editors to search for potential reviewers across the Web of Science. Over half of all journal reviewers agree that finding qualified peer reviewers to accept invitations is the most difficult part of the job. This tool significantly lessens the workload of the journal editor by ensuring that the reviewer has the appropriate background.

Once a paper is published, AI plays a continuing role through the use of identifiers and metadata tags, such as ORCHID and CRediT, as well as altmetrics to publish, archive, and dynamically search the manuscript. Analytical tools such as iCite and scite use AI-driven bibliometric approaches to understand published work. These tools are valuable for both researchers and publishers alike because they enable quick, accurate access to relevant data based on desired parameters.

AI Limitations

It is important to remember that AI is not a magic wand; it is merely a tool with limitations. A manual, human effort must always check the work of automation, especially when it comes to writing. Think of Microsoft Word’s spell check—even though it is a useful reference, it is only meant to find the most obvious errors, and a person must always read through the article. After all, while we sometimes write for machines, the actual content of the research is intended for humans. Continual improvement of the publishing process will occur as experimentation with AI in this area becomes more understood.