Policy on the Use of Artificial Intelligence (AI) Tools

JOSTA’s requirements for AI disclosure, explanatory illustrations, scientific figures and confidential editorial screening.

The Journal of Sustainable Technology in Agriculture (JOSTA) supports responsible, transparent use of artificial intelligence (AI) while protecting scientific integrity, originality, confidentiality and human accountability. This policy applies to authors, reviewers and editors, and to manuscripts, revisions and supplementary materials submitted to the journal.

Authorship and responsibility

AI tools cannot be authors or co-authors. Human authors remain responsible for the accuracy, originality and integrity of the entire manuscript, including AI-assisted content. Authors must verify facts, references, calculations, code and illustrations against reliable sources and original research records. Declaring AI use does not reduce this responsibility.

Permitted assistance

With appropriate disclosure and human verification, AI tools may assist with language correction, translation, readability, organisation, formatting, literature discovery and computer code. Authors must consult and verify the original sources they cite, test AI-assisted code and validate analytical results. AI must not replace the authors’ scientific reasoning, interpretation or critical evaluation.

Explanatory illustrations: permitted within strict limits

AI-generated or AI-assisted illustrations are permitted only to explain a concept or process; they must not constitute scientific evidence or represent study data or results. Examples include conceptual flowcharts, algorithm workflow diagrams and explanatory schematics based on content supplied and verified by the authors.

Each permitted illustration must:

• Accurately represent the concept or process described, with every label, symbol and connection checked by the authors.
• Contain no invented observations, numerical findings or unsupported scientific relationships.
• Be clearly distinguishable from experimental, observational or analytical evidence.
• Identify the AI tool in its caption and state that it is explanatory and does not represent research data or results.
• Respect third-party rights and have terms permitting its publication under the article’s licence.

Example caption:

AI-assisted explanatory illustration prepared using [tool and model/version, where available; date]. The authors verified all content. This illustration is conceptual and does not represent research data or results.

The function of an illustration determines whether it is permitted. Calling an image a “schematic” or “flowchart” does not create an exception. A study-selection or participant-flow diagram containing actual counts, exclusions or study outcomes is a study figure and is not covered by this permission.

Scientific figures and study images: generative AI prohibited

Generative AI must not be used to create, reconstruct, replace, enhance or alter scientific figures, study figures or images presented as research evidence. This includes experimental photographs, microscopy and medical images, gels, blots, field observations, and graphical representations of study results. AI must not add, remove, fill in or change evidentiary features.

Invented plots, fitted curves, statistical distributions, simulation results, data tables and numerical findings are prohibited. Authors must retain original data and source files supporting research figures.

Figures produced from verified data using reproducible plotting software, such as R or Python, are permitted. AI assistance with plotting code must be disclosed, and authors must validate the code and confirm that the figure faithfully represents the underlying data. This does not permit a generative image model to draw or recreate results.

AI used as a research method

Legitimate AI and machine-learning research must be described in the Methods section. Authors must report the tools and model versions, relevant settings, data sources, training and evaluation procedures, validation and limitations, and provide code or access information where feasible.

Synthetic data used in methodological research must be explicitly labelled as synthetic, with its generation and validation explained; it must never be represented as collected observations. Studies investigating generative image outputs must identify those outputs as objects of investigation and obtain editorial assessment before including them. Such use does not permit fabricated evidence or undisclosed alteration of study figures.

Author declaration and disclosure

Authors must disclose AI use through JOSTA’s existing author declaration signed at submission. A separate signed AI declaration is not required. The manuscript must include a statement headed “Declaration of any AI tool”, before the References, consistent with the submitted declaration and the actual use of AI. Authors must update their disclosures if AI use changes during revision.

For language assistance only, authors may use the wording already used in JOSTA articles:

Artificial intelligence (AI) tools were used solely for grammatical correction and language improvement. The authors reviewed and verified all content and take full responsibility for the accuracy and integrity of the manuscript.

Where AI was also used for permitted illustrations, coding or other assistance, the statement must describe those uses and identify the relevant tools and model versions where available. Authors must not use “solely for grammatical correction and language improvement” if AI served other purposes. Permitted illustrations also require a disclosure in each relevant caption; AI used as a research method must additionally be described in the Methods section.

Example for language assistance and a permitted illustration:

The authors used [tool and model/version, where available] for language improvement and to prepare the explanatory workflow in Figure [number]. The workflow is conceptual and does not represent study data or results. The authors reviewed and verified all AI-assisted content and take full responsibility for the accuracy and integrity of the manuscript.

Where no generative AI was used:

The authors declare that no generative AI tools were used in preparing this manuscript.

Routine non-generative spell-checking and conventional reference formatting do not require disclosure. Authors should retain relevant prompts, outputs, source files and verification records associated with substantive AI use and provide relevant records if requested by the editor, subject to confidentiality obligations.

Prohibited practices

AI must not be used to fabricate or falsify data, references, analyses or findings; conceal plagiarism; evade integrity checks through paraphrasing or “humanising”; misrepresent authorship or provenance; or introduce unchecked claims. Confidential, personal, sensitive or third-party material must not be uploaded to AI services without appropriate permission and safeguards.

Preliminary editorial screening

JOSTA uses advanced models from OpenAI and Anthropic (Claude) to assist its preliminary audit of possible AI-tool use and related manuscript-integrity concerns. These tools may highlight inconsistencies, questionable references or possible undisclosed assistance that require further examination.

AI-generated assessments and detection scores cannot reliably establish AI authorship and are not proof of misconduct. A human editor assesses concerns against the manuscript and supporting evidence. Authors may be asked for explanations, original data, code, sources or relevant AI-use records. An adverse decision must not rest solely on an AI-generated assessment or detection score.

Preliminary screening does not replace independent expert peer review. Human editors retain responsibility for all editorial decisions.

Confidentiality, reviewers and editors

Editorial AI processing must use journal-approved arrangements that protect unpublished content, author information and intellectual property. Before manuscript material is processed, the journal must assess the service’s confidentiality, retention, training-use and access terms. Only the minimum necessary content may be processed; identifying information must be removed where practicable. Sensitive or restricted material must not be processed without an appropriate basis and safeguards. Editors must retain an internal record of the screening tool, model/version where available, date and assessment of concerns.

Reviewers must not independently upload manuscripts, figures, data or review reports to external generative AI services. They must exercise their own scientific judgement and must not delegate manuscript evaluation or substantive review-report preparation to AI. Editors must not delegate acceptance, rejection or other editorial decisions to AI.

Non-compliance

Human editors will assess suspected non-compliance and give authors an opportunity to explain relevant concerns. Depending on the evidence and seriousness, JOSTA may request clarification or revision, reject a submission, or issue a correction, expression of concern or retraction after publication. Serious concerns may be referred to the relevant institution in accordance with applicable publication-ethics procedures.

This policy complements JOSTA’s Publication Ethics, Plagiarism Policy, Author Guidelines and Peer Review Policy. Its principles of human accountability and transparent disclosure are informed by COPE’s position on authorship and AI tools.