Artificial Intelligence (AI) Policy

Plant Science Horizons (PSH), published by Inspire Science and Tech Publisher (ISTP), supports responsible use of artificial intelligence, machine learning, large language models and AI-assisted technologies when their use is scientifically appropriate, transparent, reproducible, lawful and subject to meaningful human oversight. AI must never replace human authorship, scientific judgment, accountability or editorial responsibility.

Responsible AI across research and scholarly publishing

This policy applies to generative AI systems, large language models, AI agents, AI-assisted writing tools, AI image-generation or image-editing systems, machine-learning models, predictive systems and other AI-enabled tools used in research, manuscript preparation, peer review, editorial assessment or publication workflows.

The policy distinguishes between AI used to prepare a manuscript and AI used as part of the research methodology. These uses have different reporting requirements.

Human oversight is mandatory

Authorship requires human accountability

AI systems, chatbots, language models, agents and other software tools cannot be listed as authors or co-authors and should not be credited as authors in references.

Authorship requires the ability to approve the submitted and final versions, accept responsibility for the work, disclose conflicts, respond to questions about integrity, and enter into publication agreements—responsibilities that AI systems cannot assume.

Permitted as support, not as a substitute for scholarship

Appropriate uses may include:

• Improving language, readability or organization of author-generated text.
• Assisting with structured outlines, tables or non-scientific formatting.
• Supporting translation when the output is checked by a person competent in both the language and scientific subject.
• Helping authors identify questions to verify against authoritative literature.
• Supporting code development, documentation or troubleshooting when the resulting code is independently validated.

When manuscript-preparation AI must be declared

Authors must disclose generative AI or AI-assisted use that substantively creates, restructures, paraphrases, translates, summarizes, selects, or edits content intended for the manuscript.

Routine spelling, punctuation and basic grammar correction that does not substantively generate or reorganize content does not require disclosure. Conventional non-generative reference-manager functions likewise do not require disclosure.

A standardized declaration in the final manuscript

PSH may require a Generative AI Statement in the declarations section of accepted manuscripts. When reportable AI was used, the statement should identify the tool/service and explain its purpose.

Research use belongs in the Methods section

When AI, machine learning, computer vision, large language models or related systems are used as part of study design, data processing, image analysis, prediction, classification, modeling, coding or interpretation, the use must be described as part of the research methodology.

A general AI declaration does not replace methodological reporting.

Minimum information for AI-based research

Where relevant, authors should report:

Prompts, settings and provenance may be part of the method

When a generative model itself is the research instrument, authors should report enough information to understand and evaluate the workflow. Where relevant, this includes the model/version, access date, prompt or prompt template, system instructions, retrieval sources, key generation settings, number of runs, human-evaluation procedure and how outputs were selected or excluded.

Prompts and representative outputs should be archived as supplementary material or in a repository when this is feasible and legally permitted.

AI performance claims must be methodologically defensible

• Training and evaluation data must be appropriately separated.
• Authors should address data leakage, duplicate samples and information leakage across splits.
• Performance should be compared with appropriate baselines or established methods.
• External validation is strongly encouraged for models intended to generalize across genotypes, environments, species, seasons or imaging platforms.
• Limitations, domain shift and uncertainty should be discussed rather than hidden behind aggregate performance metrics.

Biological meaning matters as much as predictive performance

AI studies in plant phenotyping, disease detection, crop prediction, genomics, breeding, remote sensing, image analysis or multi-omics should describe the biological materials, genotypes, environments and sampling framework sufficiently to judge generalizability.

A high-performing model alone is not evidence of a biological mechanism. Mechanistic claims require independent biological support.

Generated code must be treated as unverified code

Authors may use AI tools to assist with code development, debugging or documentation, but they remain responsible for validating the code, checking security and logic, confirming package and function behavior, and ensuring that computational results are reproducible.

When AI-assisted code materially forms part of the research workflow, this should be disclosed in the Methods section.

AI output is not a substitute for reading the source

AI tools may assist authors in locating or organizing literature, but every cited reference must be independently verified against the original scholarly source. Authors must confirm authorship, title, journal, year, volume, pages/article number and DOI where applicable.

Fabricated, nonexistent, misattributed or materially misrepresented references may result in editorial rejection or post-publication action. Authors should cite the original source supporting a claim rather than citing an AI chatbot merely because it supplied the information.

AI-assisted translation requires verification

Generative AI may be used to translate author-generated text, but translation must be reviewed for scientific accuracy, completeness, terminology, bias and unintended changes in meaning by a person competent to evaluate the source and target language.

Generative AI-assisted translation included in the manuscript is a reportable use.

AI must never fabricate scientific evidence

Prohibited uses include:

• Inventing experimental measurements, biological observations or sample metadata.
• Generating false results to fill missing experiments or replicates.
• Creating fabricated statistical outputs or model performance values.
• Generating nonexistent references, quotations, ethics approvals, permits or accession numbers.
• Using AI to conceal plagiarism, duplicate publication or manipulation of the research record.

No generative creation or alteration of data-bearing images

Generative AI must not create, replace, extend, reconstruct or alter images that represent primary research evidence, including microscopy, histology, gels, western blots, phenotyping photographs, field images, disease symptoms, tissue images or other directly observed experimental material.

AI must not add, remove or transform biological features in a way that changes the underlying evidence. Permissible conventional adjustments must comply with the PSH image-integrity policy and be applied consistently without changing interpretation.

Visual output must remain traceable to real data

AI tools may support the generation of plots, heatmaps, diagrams or other data visualizations only when the output is faithfully derived from the underlying dataset through a reproducible analytical or computational workflow.

AI must not invent data points, interpolate unsupported values, change sample identities or alter the visual presentation in a way that misrepresents the analysis.

Non-data AI-assisted illustrations require transparency

AI-assisted conceptual or explanatory illustrations may be considered when they do not represent primary research evidence. Examples include a schematic mechanism, conceptual workflow or educational illustration.

Such use must be disclosed in the figure caption and Generative AI Statement, including the tool and its role. Authors must verify scientific accuracy, originality and rights compliance and must not create misleading photorealistic representations of observations that did not occur. Editors may require replacement with an author-created schematic when provenance, accuracy or copyright status is uncertain.

Authors must have the right to upload and publish what they use

Before using an AI service, authors should consider whether uploaded text, data, images or code may be retained, reused, shared or used for model training. Authors must not upload confidential, proprietary, personally identifiable or third-party material unless they have authority to do so and the tool provides appropriate protections.

Authors remain responsible for ensuring that AI-generated or AI-assisted outputs do not infringe copyright, privacy, database rights, licenses, contracts or other legal rights.

Authors may be asked to document material AI use

For substantive generative AI use, authors are encouraged to retain a reasonable record of the tool, model/version, purpose, relevant prompts or instructions, generated output and how the authors reviewed or modified that output.

PSH may request such documentation when necessary to evaluate transparency, provenance, reproducibility or a research-integrity concern.

Do not upload confidential manuscripts to public or unapproved AI systems

A manuscript under review is confidential. Reviewers must not upload the manuscript, supplementary files, unpublished data, figures, reviewer correspondence or identifiable manuscript excerpts to public, consumer or otherwise unapproved AI services.

This restriction also applies to external AI-detector services when uploading manuscript content would disclose confidential material.

AI cannot perform the substantive peer review

Reviewers must personally evaluate the manuscript and remain responsible for every statement in the report. AI-generated review reports must not substitute for independent expert assessment.

If PSH later permits a secure, privacy-preserving AI tool for limited review assistance, any use beyond basic grammar or formatting should be disclosed to the editor, and the reviewer remains fully accountable for the final report.

Editorial responsibility remains human

Editors and journal staff must protect confidential manuscript material and should not upload unpublished submissions to public or unapproved generative AI systems.

Secure publisher-controlled tools may support administrative tasks, metadata checking, similarity screening, workflow triage or integrity assessment when confidentiality, data protection and appropriate human oversight are maintained. AI may assist a process; it may not make the final editorial decision.

Detector scores are not proof of misconduct

AI-text or AI-image detection tools may produce false positives and false negatives and should not be treated as definitive evidence that content was or was not generated by AI.

PSH will not base an allegation of misconduct or an editorial sanction solely on an automated AI-detector score. Suspected undisclosed or prohibited AI use should be assessed using the manuscript, provenance information, source data, author explanation and other relevant evidence.

Assess concerns neutrally and evidence-first

If editors have credible concerns that reportable or prohibited AI use was not disclosed, the authors may be contacted for clarification and asked to provide relevant information or source material.

Suspicion alone should not be represented as established misconduct. Editorial action should be proportionate to the evidence and effect on the reliability of the work.

Consequences depend on seriousness

Failure to disclose permissible AI assistance may require clarification or correction. More serious cases—such as fabricated references, generated experimental evidence, falsified data, deceptive images, plagiarism concealment or compromised peer review—may lead to rejection, withdrawal of consideration, institutional referral, correction, expression of concern or retraction as appropriate.

The journal's Research Integrity and Retraction policies govern formal post-publication action.

AI cannot replace the authors' intellectual contribution

PSH does not accept manuscripts that are substantially generated by AI without genuine human intellectual contribution, critical verification and transparent disclosure.

Authors must be able to explain and defend the reasoning, evidence, methods, references and conclusions in the submitted work.

Accessibility support should not be treated as misconduct

Assistive technologies used solely for accessibility, spelling, punctuation or other non-substantive support are not treated as substantive authorship by AI. Authors remain responsible for ensuring that any resulting manuscript content is accurate.

Related author, editorial and publication guidance

Evolving publication-ethics standards

This policy is informed by recognized scholarly-publishing guidance concerning AI authorship, transparency, confidentiality, peer review and research integrity, including relevant guidance and discussions from the Committee on Publication Ethics (COPE) and other established scholarly-publishing organizations.

Reference to external guidance does not imply membership, certification, accreditation or endorsement unless separately confirmed.

AI policy will evolve with technology and publishing standards

PSH may revise this policy as AI technologies, privacy protections, copyright standards, research practices and scholarly-publishing guidance evolve. Authors, reviewers and editors should consult the current online version when using AI in connection with a PSH submission.

What PSH expects now

AI may assist; humans remain accountable

Plant Science Horizons supports responsible innovation while preserving the foundations of scholarly communication: authentic human intellectual contribution, reproducible research, truthful evidence, transparent disclosure, confidential peer review and accountable editorial judgment. AI tools may support these activities, but they cannot replace the people responsible for the science or the publication decision.