Responsible Research

AI in Academic Publishing: Acceptable Uses, Risks, and Disclosure

Speaker

Prof. Roger Watson

Prof. Roger Watson42 min

Quick answer

When using SciSpace in academic publishing, AI may support search, editing, analysis, figures, or manuscript preparation, but authors remain accountable for accuracy, originality, confidentiality, and disclosure. Check the target journal’s policy, document the tool and purpose, verify all generated content, and never assign authorship or reviewer responsibility to AI.

About this session

Prof. Roger Watson outlines the impact of AI on academic publishing, focusing on its role in automating submission checks and streamlining editorial workflows [L5-L10]. The discussion highlights the use of AI-based systems for plagiarism detection while emphasizing current editorial limitations on AI usage by authors and reviewers during peer review [L10-L16].

AI is increasingly used to search, edit, summarize, and support manuscript preparation, but editorial responsibility cannot be delegated to a tool. This expert session examines where AI assistance may be acceptable, where it creates ethical or practical risks, and what authors should document before submission.

What this tutorial covers

Distinguish language or workflow assistance from authorship, scientific judgment, and accountability.
Recognize risks involving fabricated citations, altered meaning, plagiarism, bias, and confidential material.
Interpret journal and publisher guidance rather than assuming one universal AI policy.
Prepare a transparent disclosure that describes the tool, purpose, stage, and human verification.

Key concepts explained

Authorship requires accountability: An AI system cannot approve the manuscript or answer for its integrity.
Editorial confidentiality: Unpublished manuscripts and peer-review material may be restricted from external tools.
Disclosure specificity: Useful statements explain what the tool did, not simply that 'AI was used.'.
Policy hierarchy: Journal, publisher, institution, funder, and disciplinary rules may all apply.

How to try it in SciSpace

1
List every stage where AI was used, including search, analysis, writing, editing, figures, or review.
Create an inventory of every stage where AI assisted: search, screening, analysis, coding, writing, editing, translation, figures, or review. Record the tool and purpose for each stage.
2
Check the current policy of the target journal, publisher, institution, and relevant reporting guideline.
Check the current journal, publisher, institution, funder, and reporting-guideline policies. Save the URL, version or date, and any conflicting requirements.
3
Verify all generated claims, citations, quotations, and interpretations against primary sources or study data.
Verify generated claims, citations, quotations, statistics, and interpretations against primary sources or the study record. Authors remain responsible even when the text is polished.
4
Remove confidential or personally identifiable information unless approved safeguards are in place.
Remove confidential manuscripts, peer-review content, personal data, or restricted material unless the chosen tool and workflow are explicitly approved for them.
5
Draft a precise AI-use disclosure and retain prompts or logs when appropriate.
Draft a specific disclosure that names the tool, task, stage, output, material changes, and human verification. Retain prompts or logs when policy or reproducibility requires them.

Worked example

An author uses SciSpace to create a literature matrix and improve the language of a manuscript but does not use it to generate data or make analytic decisions. Before submission, the author checks the journal’s AI policy, verifies every edited claim and reference, and drafts a disclosure that names the tool, the tasks performed, and the author’s review. The disclosure avoids claiming that AI is an author or that it takes responsibility for the work.

Practice prompt

Audit my manuscript workflow for AI use. For each stage, classify the activity, identify authorship, confidentiality, citation, bias, and reproducibility risks, state what human verification is required, and draft a factual disclosure aligned with the target journal’s current policy.

Key takeaway

Editors may accept AI-assisted work, but authors remain fully responsible for accuracy, originality, disclosure, confidentiality, and the final scientific argument.