Does AI-Assisted Translation Need to Be Disclosed in Research?

If AI Translates Your Research, Must You Disclose That Assistance?

A researcher writes a manuscript in Filipino and uses an AI translation tool to prepare an English version for journal submission. The researcher reviews the translated text and corrects several technical expressions before submitting the paper.

In another study, a researcher uses AI to translate interview transcripts from Filipino into English before conducting qualitative analysis.

Both activities involve translation, but they affect different parts of the research process. The first concerns how scholarship is communicated, while the second may influence how research evidence is interpreted.

Should both forms of assistance be disclosed? Does the answer change if the researcher uses DeepL rather than ChatGPT? What if the translation is accurate but the journal does not explicitly mention translation in its AI policy?

These questions require distinguishing the technological assistance from its methodological consequences and the reporting requirements that apply.

AI Translation May Require Disclosure, Especially When It Affects Research Data

The most consequential distinction is not which translation application was used, but whether the translation changed only the language of presentation or formed part of the procedures through which research evidence was produced.

When AI Translation Becomes a Disclosure and Methodological Issue

Why Is Translation Different From Ordinary Grammar Correction?

Grammar correction generally modifies existing language, while translation transfers meaning between languages.

That process can involve choices concerning terminology, cultural context, emphasis, and interpretation. Words that appear equivalent in a dictionary may carry different meanings in a particular academic discipline or research setting.

For example, translating the Filipino expression "pakikisama" simply as "cooperation" may overlook aspects of interpersonal accommodation and maintaining harmonious social relationships that could be relevant to a qualitative analysis.

Likewise, technical terms in educational research may have established meanings that are not preserved by literal translation.

AI translation can therefore introduce interpretive changes even when the resulting English text is grammatically fluent.

The distinction matters because AI-assisted grammar correction and AI-assisted translation may not fall under identical publisher requirements.

Does Translating a Research Manuscript Require Disclosure?

It depends on the intended journal's policy.

Some publishers require disclosure of generative AI assistance in manuscript preparation, while others distinguish routine language editing from substantive content generation.

Translation may fall within a publisher's language-assistance rules, but researchers should not assume that an exemption for grammar correction automatically extends to full manuscript translation.

For example, translating an independently written introduction from Filipino into English may involve extensive reformulation of sentences without necessarily changing the underlying argument.

However, a generative AI system may also add explanations, strengthen claims, or alter technical terminology during translation.

The researcher should examine the translated output and determine whether the assistance remained faithful to the original or introduced substantive content.

Where disclosure is required, the statement should identify the translation assistance accurately.

What Do ICMJE and COPE Say About AI Translation?

The International Committee of Medical Journal Editors (ICMJE) recommends disclosure of AI-assisted technologies used in producing submitted work and emphasizes human responsibility for the resulting content.

Its guidance distinguishes AI writing assistance from AI use in research methods.

Although its general AI recommendations do not establish a separate universal rule for every machine-translation activity, they provide a framework for evaluating translation used during manuscript preparation or research.

The Committee on Publication Ethics (COPE) similarly emphasizes transparency about AI tools used in manuscript writing and research activities.

Neither position should be interpreted as establishing that every translation performed by every software application must be declared in exactly the same way.

Researchers should consult the journal's implementation of these policies and distinguish manuscript translation from translation performed as part of the research methodology.

Does Springer Nature Treat AI Translation as Copyediting?

Springer Nature distinguishes defined AI-assisted copyediting from substantive generative AI use.

Its copyediting exemption concerns specified assistance with human-generated text, including improvements to readability, style, grammar, spelling, punctuation, and tone.

That exemption should not automatically be interpreted as covering every form of translation, particularly when an AI system reconstructs substantial passages or introduces new content.

Researchers should examine the current publisher guidance and the intended journal's instructions to determine whether the particular translation activity falls within an exempted category.

The relevant consideration is the actual function performed, not simply whether the researcher describes it as language assistance.

Does the Translation Tool Matter?

The tool may matter because different applications perform different functions, but its name alone does not determine the disclosure requirement.

DeepL, Google Translate, ChatGPT, and other services may offer machine translation or generative rewriting capabilities.

Some tools primarily translate supplied text, while others may also paraphrase, expand, summarize, or restructure it.

The applicable journal may distinguish basic translation utilities from generative AI systems, or it may regulate AI-assisted language processing more broadly.

Researchers should identify the specific functionality used rather than assuming that every feature of a particular application has the same disclosure status.

They should also verify the service's data-handling arrangements before uploading unpublished manuscripts or confidential research materials.

When Does AI Translation Become Part of the Research Methodology?

Translation becomes methodologically relevant when it affects the collection, processing, analysis, or interpretation of research evidence.

Examples include translating questionnaires before administration, translating participant interviews for coding, and translating documentary sources used in content analysis.

In these circumstances, the translation is not merely a presentation aid. It may influence what participants understand, how their responses are categorized, or how researchers interpret the evidence.

Methodological reporting should therefore explain the translation procedure sufficiently for readers to evaluate its implications.

A brief acknowledgment that AI was used may not provide the necessary information.

The distinction between methods reporting and acknowledgment of AI assistance is particularly important for multilingual research.

Should AI Translation of Interview Transcripts Be Disclosed?

AI translation of interview transcripts generally warrants methodological transparency when the translated material is used for analysis.

Qualitative interviews often contain expressions whose meaning depends on cultural context, tone, and the circumstances of the conversation.

Automated translation may preserve the general topic while losing distinctions important to interpretation.

For example, a participant's expression of uncertainty may be translated into a more definite statement, affecting how the response is coded.

Researchers should describe how transcripts were translated, whether original-language materials were retained, and how translation accuracy was evaluated.

Depending on the study, bilingual review, comparison with original excerpts, or other validation procedures may be appropriate.

Such reporting is relevant even when the AI-generated translation does not appear verbatim in the published manuscript.

Does Translating Research Instruments Require Disclosure?

Translation of questionnaires, interview guides, and other research instruments may affect measurement validity.

An item translated into another language may not preserve the same conceptual meaning or level of difficulty.

For example, an English questionnaire item measuring perceived usefulness may acquire a different nuance when translated into Filipino.

If AI is used to prepare the translated instrument, the researcher should evaluate whether the wording is conceptually and culturally appropriate.

Depending on the study, expert review, cognitive interviewing, pilot testing, or other validation procedures may be necessary.

Methodological reporting should explain the relevant translation and validation procedures.

AI translation does not eliminate the need to establish that the instrument measures the intended construct in the target language.

Can AI Translation Affect Qualitative Coding?

Yes. Translation can influence the categories and themes identified during qualitative analysis.

Suppose a researcher analyzes translated interview transcripts rather than the original-language responses.

If the AI translation changes the emotional tone or meaning of an expression, subsequent coding may reflect the translated wording rather than the participant's intended meaning.

This risk is especially relevant when concepts do not have direct equivalents across languages.

Researchers should consider whether coding should occur in the original language, the translated language, or through a documented combination of both.

Where AI translation materially influences the analytical process, the methods section should describe that involvement and the verification undertaken.

These considerations are closely related to disclosing AI-assisted qualitative analysis.

What About Translating Quotations for Publication?

Translated participant quotations can affect how readers understand qualitative findings.

A quotation selected to illustrate a theme should represent the participant's meaning accurately.

If AI assists with translating quotations, researchers should verify the translation against the original response, particularly where cultural expressions or ambiguous wording are involved.

Depending on the journal and research design, it may be appropriate to explain the translation procedure in the methods section or a note accompanying the quotations.

Researchers should avoid presenting an AI-generated paraphrase as a direct translation when it materially changes the original statement.

Translation transparency is especially important when readers cannot independently evaluate the source-language material.

Does AI Translation of Published Literature Require Disclosure?

AI translation may assist researchers in reading publications written in languages they do not understand fluently.

Whether that assistance requires disclosure depends on the journal's policy and the role of the translated material in the research.

For example, using a translation tool to understand an individual article during preliminary reading may be treated differently from systematically translating and extracting data from multilingual studies for a review.

When translation forms part of a documented evidence-extraction procedure, methodological reporting may be necessary.

Researchers should also verify important claims against the original source or obtain appropriate language expertise when translation uncertainty could affect the synthesis.

A fluent machine translation does not guarantee accurate representation of the source publication.

Does AI-Assisted Translation Require Human Verification?

Yes, particularly when translation affects substantive scholarly claims or research data.

The appropriate level of verification depends on the importance and consequences of the translated material.

For manuscript prose, researchers should check that technical terminology, uncertainty, and intended meaning are preserved.

For research instruments, verification may need to address conceptual equivalence and measurement validity.

For qualitative data, researchers should examine contextual meaning and possible interpretive changes.

Verification may involve bilingual researchers, qualified translators, comparison with original-language materials, or other procedures appropriate to the study.

Human review should not be described as comprehensive bilingual validation unless that process actually occurred.

Can AI Translation Introduce Bias Into Research?

Potentially. Translation systems may perform unevenly across languages, dialects, registers, and culturally specific expressions.

They may favor more standardized formulations or resolve ambiguous statements in ways that alter the original meaning.

In qualitative research, such changes can affect how participants' experiences are represented.

In survey research, they may affect how respondents understand questionnaire items.

These risks do not mean that AI translation is necessarily inappropriate. They indicate that its suitability and verification should be assessed according to the research context.

Researchers should avoid assuming that grammatical fluency in the target language establishes semantic or cultural equivalence.

What About Confidentiality When Translating Research Data?

Uploading participant interviews, unpublished manuscripts, or other confidential material to an external translation service may create privacy and research ethics risks.

The service may process or retain information under terms that are incompatible with the study's consent arrangements or institutional requirements.

Researchers should determine whether the intended tool is authorized for the data involved before uploading any material.

Where necessary, they may need an institutionally approved service, appropriate de-identification, or another translation arrangement.

De-identification alone may not eliminate all re-identification risks, particularly for detailed qualitative narratives.

Disclosure after publication cannot retrospectively authorize a data transfer that violated applicable obligations.

Where Should AI-Assisted Translation Be Disclosed?

The reporting location depends on the role of translation and the journal's instructions.

These locations are illustrative rather than universal. The intended journal may require a separate declaration or additional details.

What Should an AI Translation Disclosure Include?

A useful disclosure should identify the tool, the purpose of translation, the materials affected, and the verification performed, as required by the journal.

For methodological translation, researchers may also need to explain the languages involved, the stage of analysis at which translation occurred, and how accuracy was evaluated.

This example assumes that the stated verification procedures actually occurred. Researchers should name the tool and provide further details where required by the journal.

The statement should not claim that translations were independently validated if no such validation took place.

Does Passing an AI Detector Eliminate Translation Disclosure Requirements?

No. AI-text detectors do not determine whether translation assistance must be disclosed.

A translated manuscript may receive a low or high AI-detection score for reasons unrelated to the actual translation process.

Disclosure requirements depend on the assistance used and the applicable policy.

Researchers should not treat detector results as evidence that translation was unnecessary to report or that the resulting manuscript was written without AI assistance.

What If the Journal Does Not Mention AI Translation?

Researchers should review the publisher's broader AI-use policy, language-editing instructions, and submission declarations.

For translation affecting research methods, the ordinary requirement to describe consequential procedures may remain relevant even without a specific AI-translation rule.

If uncertainty remains, the researcher may request clarification from the editorial office.

The inquiry should describe whether AI translated manuscript prose, research instruments, participant data, or other materials.

These distinctions are more informative than asking whether the journal permits machine translation in general.

When AI Translation Changes the Interpretation of an Interview Response

The example illustrates how translation can affect research interpretation even when the translated sentence is grammatically correct.

The methodological issue is not simply whether AI was used, but whether the translated material preserves the meaning required for credible analysis.

Disclosure helps readers understand the process, while verification helps protect the integrity of the findings.

Misconceptions About AI Translation and Research Disclosure

How Should You Decide Whether to Disclose AI Translation?

Begin by identifying what was translated and whether the translated material influenced the research process or only its presentation.

Then determine the applicable policy and the verification necessary to preserve meaning.

Researchers should avoid treating translation as a purely technical activity when it can influence the meaning of evidence.

The appropriate standard is faithful translation, proportionate methodological transparency, and compliance with the relevant publication and research ethics requirements.

What to Verify Before Using or Reporting AI Translation

Frequently Asked Questions About AI-Assisted Translation Disclosure

AI Translation Requires Attention to Both Disclosure and Meaning

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.