About AI-JISR How It Works →

Our Approach

Academic publishing is one of the primary mechanisms for evaluating, disseminating, and building upon scientific knowledge. However, growing submission volumes, difficulties in recruiting reviewers, editorial workload, and limited review capacity can impede this mechanism. Publication delays affect the visibility of researchers' academic work and the timely circulation of findings. The starting point of our journal is to respond to these challenges with a more predictable, traceable, and clearly defined evaluation model.

Long waiting times have been addressed in the literature with concrete data. In a systematic review of 69 biomedical publishing studies, Andersen, Fonnes, and Rosenberg (2021) found that reported average submission-to-publication times ranged from 91 to 639 days. Horta and Jung (2024) discussed pressures on the peer review system in the context of growing publication output, researcher workload, the voluntary nature of reviewing, and review capacity.

Current findings on the difficulty of recruiting reviewers further support this discussion. In a study examining records from Estuaries and Coasts for 2007–2025, Montagna (2026) found that the average number of invitations needed to obtain two reviewer reports had doubled, from four to eight.

AI has moved beyond an abstract possibility added to this discussion and has begun to be used in peer review practices. Liang et al. (2024) estimated that approximately 6.5–16.9% of review texts from leading AI conferences may have been substantially produced or modified by large language models. A corporate report by Frontiers (2025) based on a survey reported that 53% of active researchers who conduct peer review said they use AI tools in that role.

The contributions of AI to editorial processes have also been examined across specific tasks. In a study of 20 journal editors evaluating AI-suggested reviewers, Farber (2024) showed that reviewer selection tools save time and help identify suitable researchers who might not otherwise have been considered. More recently, in a randomized controlled trial published in Nature Machine Intelligence covering over 20,000 reviewer reports at ICLR 2025, Thakkar et al. (2026) found that 27% of reviewers who received automated feedback revised their reports, resulting in more informative reviews.

Editorial Model

In our journal, the assistant editor, field editors, and all reviewers are AI agents. This structure is disclosed to authors before submission.

• Assistant Editor Agent: Conducts scope review and routes submissions to the appropriate field.
• Field Editor Agents: Handle field-level evaluation, assignment of reviewer agents, and synthesis of reports.
• Reviewer Agents: Assess manuscripts according to the journal's published scientific criteria.
• Editor-in-Chief (Human): Initiates the process, monitors the workflow, oversees outcomes, and bears responsibility for final publication decisions.

The agents were configured specifically according to the academic topics within the journal's scope and the tasks they undertake. During this development process, highly ranked journals in relevant fields on SCImago Journal & Country Rank, high-quality scientific publications, reviewing criteria, and editorial evaluation practices were surveyed as references.

AI processes are conducted autonomously using large language models accessed through the NVIDIA API. Models are used with task definitions appropriate to agent roles and structured output formats.

Reliability and Ethics

We do not regard reliability as an assumed property based on the name of the technical infrastructure or the speed of the model. Perlis et al. (2025) drew attention to hallucination, bias, privacy, and the preservation of human accountability when discussing AI use in peer review. A study by Li et al. (2024) examining the policies of the top 100 medical journals showed that data privacy is a primary concern.

For this reason, our journal adopts the following principles:

• Evaluation rationales must be inspectable
• Errors can be corrected and authors have the right to appeal
• The editor-in-chief is accountable
• Personal data and unpublished research content are kept confidential

Transparency

Publicly disclosed level: Agent roles, evaluation criteria, scoring approach, editorial workflow, models used, and publication policies.

Article-specific level: Processing history, reviewer agent reports, revision requests, author responses, and decision rationales — made available for review by the author and editor-in-chief.

A core principle is that authors are informed before submission about the technical infrastructure through which their manuscript will be processed and the scope of AI use.

Publication Capacity and Development

Each issue accommodates a maximum of 10 articles. This limit reflects a preference for managing publication capacity and ensuring that each study receives appropriate attention during the evaluation process.

The development of agents is an ongoing process. As emphasised by Aczel et al. (2025), fair assessment of changes in peer review requires more comprehensive and rigorous data.

Submission to our journal is voluntary. We are open to researchers who wish to examine our approach and contribute to this publishing experience. Our journal is a developing publishing proposal; it welcomes critique as one of the scientific foundations of that development.

Academic References