Valence
VALENCE: Seven Rules for Responsible AI Use
What VALENCE allows
- language and stylistic editing, translation, and shortening of text;
- drafting alternative titles, abstracts, structures, and transitions between sections;
- formulating self-check questions and identifying unclear passages in a manuscript;
- preliminary organization of the literature and preparation of working summaries — with mandatory verification of the primary sources;
- other auxiliary tasks, provided the researcher understands the tool’s limitations and verifies the output.
What VALENCE does not allow
- generating non-existent data, experiments, observations, or references and presenting them as scientific results;
- accepting a machine interpretation of data as a scientific conclusion without specialist verification;
- concealing material use of AI when the journal or institution requires disclosure;
- shifting responsibility for errors, plagiarism, rights violations, or unreliable conclusions onto AI;
- uploading confidential or unpublished materials to external services without checking security terms and institutional policy.
Why these rules apply beyond chemistry
VALENCE emerged from the editorial practice of chemistry journals, but its seven principles do not depend on a particular research object. Scientists in every field need fact checking, transparency about tool use, protection of sources, and preservation of human responsibility.
At the same time, different fields of science should supplement VALENCE with discipline-specific requirements. VALENCE is therefore a foundational layer of governance that can be adapted to the professional standards of each discipline.
How to read VALENCE
V — Verify
Verify facts, names, formulas, units, numbers, references, data, and conclusions. A convincing statement proposed by AI is not evidence.
Self-check: can I support every material statement with primary data, a reliable source, or a reproducible calculation?
A — Acknowledge
Openly disclose material use of AI — especially when it helped create, translate, edit, structure, or analyze text.
Self-check: will the reader understand which AI tool was used, for what purpose, and where its contribution was material?
L — Limit
Use AI where its performance can be checked: language, structure, drafting, identifying possible errors, and preliminary organization of information. Do not delegate scientific judgment to a machine in place of a specialist.
Self-check: have I delegated to AI a decision that a researcher should make on the basis of data and professional expertise?
E — Edit
A draft produced with AI assistance must undergo full authorial revision. The final text must reflect real data, reasoning, knowledge, and the researcher’s own scientific position.
Self-check: does the final version reflect my scientific position, and can I defend every material part of it?
N — No AI authorship
AI, a chatbot, a language model, or a software service cannot be an author or co-author of a scientific work: it cannot accept responsibility, confirm a contribution, or be accountable for errors.
Self-check: can the listed “author” knowingly confirm the contribution and take responsibility for the publication? If not, it is not an author.
C — Cite & protect
Cite primary sources, avoid plagiarism, verify rights to use materials, and protect data confidentiality. Use reliable services with clear terms for information processing.
Self-check: have the primary sources, rights, confidentiality terms, and permission to upload unpublished materials been checked?
E — Ensure standards
Maintain accuracy, clarity, originality, reliability, reproducibility, ethical integrity, and compliance with the requirements of the target journal and your scientific field.
Self-check: has the work improved by scientific criteria — rather than merely becoming more fluent?
For authors: how to cite VALENCE in a manuscript
First, check the policy of the target journal: it determines where and in what form the use of AI should be disclosed.
If the journal does not provide a dedicated section, this information is usually placed in a separate “Declaration of AI Use” or in the Methods section.
Citing VALENCE shows that the authors used AI within a clear framework of responsibility accepted by the scientific community; it does not replace disclosure of the specific tool and the specific task for which it was used.
Short version — ready-to-use text
During preparation of the manuscript, the authors followed the VALENCE principles [ref]. AI was used only as an auxiliary tool; all text, references, data, interpretations, and conclusions were reviewed and approved by the authors, who take full responsibility for the content of the work.
[ref] S. M. Aldoshin, V. P. Ananikov, I. P. Beletskaya et al. Artificial Intelligence in the Preparation of Scientific Articles. Russ. Chem. Rev. 2026, 95, RCR5247. DOI: 10.59761/RCR5247. DOI
Extended version — ready-to-use text
During preparation of this manuscript, the authors followed the VALENCE principles [ref] and used [tool name, version/date] for [specific task and, where appropriate, manuscript section: translation, language editing, text structuring, drafting wording, preliminary organization of the literature, etc.]. After using the tool, the authors independently verified facts, numbers, references, terminology, interpretations, and conclusions, edited and approved the final text. The AI tool was not used as an independent source of scientific conclusions, is not an author or co-author, and bears no responsibility for the content of the publication. Full responsibility for the scientific accuracy, originality, and reliability of the work remains with the authors.
[ref] S. M. Aldoshin, V. P. Ananikov, I. P. Beletskaya et al. Artificial Intelligence in the Preparation of Scientific Articles. Russ. Chem. Rev. 2026, 95, RCR5247. DOI: 10.59761/RCR5247. DOI
The bibliographic reference should be adapted to the style of the target journal.
For editors: how to incorporate VALENCE into journal policy
VALENCE can serve as a concise basis for an AI-use section in the Instructions for Authors.
Short wording for the Instructions for Authors
This journal permits responsible use of AI tools in manuscript preparation. Material use of AI must be transparently disclosed in accordance with the journal’s policies. AI cannot be an author or co-author. Authors must independently verify facts, references, data, terminology, originality of the text, and compliance with applicable rights, and they bear full responsibility for the scientific content of the publication. The editors recommend using the VALENCE principles [ref] as a general framework.
[ref] S. M. Aldoshin, V. P. Ananikov, I. P. Beletskaya et al. Artificial Intelligence in the Preparation of Scientific Articles. Russ. Chem. Rev. 2026, 95, RCR5247. DOI: 10.59761/RCR5247. DOI
Extended wording for the Instructions for Authors
Authors may use generative and other AI tools for auxiliary tasks in manuscript preparation, provided that such use complies with journal policy and the requirements of the research institution. Any AI use that materially affected the creation, translation, editing, structuring, or analysis of the text must be disclosed, including the tool and the purpose for which it was used. AI may not be listed as an author or co-author and is not a source to which scientific judgment or responsibility can be delegated. Authors must verify all material claims, bibliographic references, numbers, data, terminology, and conclusions; ensure originality of the text, appropriate citation, copyright compliance, and confidentiality. A probabilistic AI-text detector output should not be treated as standalone evidence of misconduct; editorial decisions should be based on specific issues of scientific reliability, originality, transparency, or ethics. As a general framework, the journal recommends the VALENCE principles [ref]; specific requirements of the journal and the relevant scientific field take precedence.
[ref] S. M. Aldoshin, V. P. Ananikov, I. P. Beletskaya et al. Artificial Intelligence in the Preparation of Scientific Articles. Russ. Chem. Rev. 2026, 95, RCR5247. DOI: 10.59761/RCR5247. DOI
What every researcher should remember
AI is a tool, not the bearer of scientific responsibility.
Every fact, reference, number, and conclusion must be independently verifiable, regardless of how the draft was created.
Material use of AI is better disclosed clearly and specifically than concealed.
The closer a task is to a scientific conclusion, the less autonomy should be given to AI and the greater the need for human verification.
The rules of the specific journal, institution, and discipline must always be followed.
About VALENCE
A collective editorial position of Russian scientists gives researchers the “green light” for responsible use of AI.
From a practical problem to a common language
Artificial intelligence entered scientific practice faster than the research community could establish shared rules for its use. Researchers already use AI for translation, editing, structuring text, preparing drafts, and navigating information, yet often remain unsure what should be documented, how the tool’s contribution should be described, and where the boundary lies between assistance and substitution for scientific judgment.
The VALENCE project emerged in response to this uncertainty. In 2026, Russian Chemical Reviews published the article “Artificial Intelligence in the Preparation of Scientific Articles,” which formulated seven VALENCE principles — Verify, Acknowledge, Limit, Edit, No AI authorship, Cite & protect, Ensure standards. Their overall message is simple: AI may be used, but responsibility for the scientific content always remains with the researcher.
Authors of the article
S. M. Aldoshin, V. P. Ananikov, I. P. Beletskaya, A. A. Berlin, L. B. Boinovich, V. I. Bukhtiyarov, A. A. Voshkin, Yu. G. Gorbunova, K. V. Grigorovich, U. M. Dzemilev, M. P. Egorov, I. L. Eremenko, V. K. Ivanov, S. N. Kalmykov, V. P. Kolotov, A. L. Maximov, A. M. Muzafarov, B. F. Myasoedov, O. G. Sinyashin, K. A. Solntsev, A. R. Khokhlov, A. Yu. Tsivadze, A. A. Yaroslavov, A. B. Yaroslavtsev.
Links to the article
A collective project
The article was the result of a dedicated collective effort by 24 scientists representing leading research organizations and the editorial community of Russian chemistry journals. VALENCE was developed not as a private opinion, but as an attempt to establish a shared and practically applicable set of principles and rules for authors, editors, and reviewers.
The author team includes editors-in-chief and editorial board members of Russian chemistry journals. Of the 24 scientists, 21 are Academicians of the Russian Academy of Sciences (RAS) and 3 are Corresponding Members of the RAS; two co-authors of the article are Vice-Presidents of the RAS. Together, the authors represent 13 leading research organizations.
A “green light” for competent and responsible use of AI
VALENCE is designed as a set of principles for calm, transparent, and effective use of new tools. The rules help distinguish useful automation from situations in which scientific responsibility begins to be shifted to a machine.
This approach is especially important for scientific progress: a prohibitive logic encourages people to hide the use of new technologies, whereas a clear system of disclosure and verification builds trust and allows useful tools to be adopted quickly. VALENCE promotes not fear of AI, but a culture of its professional use.
A project for all fields of science
The original article addresses the needs of the chemistry and editorial communities and deliberately focuses on preparation of the text of a scientific publication. However, the core of VALENCE — transparency, verification, limiting the role of AI, authorial revision, no AI authorship, proper citation, and adherence to standards — is universal.
The mission of VALENCE therefore extends beyond a single discipline. The project is intended to provide a clear starting point for researchers across the natural and engineering sciences, medicine and life sciences, the social sciences, and the humanities. Each field can and should develop specific additions reflecting the characteristics of its own research practice.
Where VALENCE is going
The VALENCE project is being developed as an open access point to the idea itself, practical guidance, and ready-to-use wording for authors and editors. Russian and English versions are intended to make the rules convenient for international use, while future development may include adaptations across different disciplines, educational materials, and examples of editorial policies.
The project’s central goal remains unchanged: not to slow the use of AI in science, but to make it mature, transparent, and professional. New tools should expand the researcher’s capabilities without diluting authorship, the quality of scientific work, or the scientist’s responsibility.
VALENCE is a shared agreement on the boundary of responsibility: technologies may change very quickly, but scientific integrity must remain non-negotiable.