AI Disclosure for Psychology Research: Participant Data, Stimuli, and APA Journal Submissions

Psychology papers need AI disclosure before the methods section turns vague

Psychology researchers often use AI in places readers cannot see.

A model may help draft survey items. It may summarize interview excerpts, rewrite participant-facing text, suggest R code, generate pilot stimuli, or polish the final manuscript. None of those uses turns the tool into an author. But each one can affect how readers judge the study.

That is the point of disclosure.

APA gives direct guidance for APA journal submissions: if authors use generative AI to draft a manuscript, they must disclose that use in the methods section and cite it. APA also says AI cannot appear as an author, and authors remain responsible for checking information and citations that an AI tool produces. (apa.org)

For psychology, the stakes sit close to the data. Participant privacy, informed consent, clinical sensitivity, and measurement validity all matter. A short sentence in the acknowledgments may not tell readers enough.

If you want a structured record, generate an [[AI Usage Card](/ai-disclosure-for-social-science-research/)](/) while the project is still moving. You can attach the card as supplementary material, copy its text into a disclosure statement, or use it as your lab record before submission.

What psychology researchers should record

Start with the simple facts.

Record the tool name, model or version if known, date of use, purpose, prompts, output type, and the paper section that changed because of the output. APA's AI contribution statement guide asks authors to track dates, prompts, affected sections, and how AI-generated ideas entered the paper. (apastyle.apa.org)

That sounds bureaucratic until you try to reconstruct AI use six months later.

You do not need to save every minor autocomplete event. You should record AI use that affected research design, materials, analysis, interpretation, writing, tables, images, code, or peer review. The more the AI shaped the research object, the more detail readers need.

A psychology disclosure record can include:

• AI-assisted literature search terms
• Drafted interview prompts or survey items
• Generated stimulus text, images, or vignettes
• Translation or simplification of participant-facing materials
• Qualitative coding suggestions
• Statistical code suggestions
• Manuscript drafting, editing, or reference checking
• Reviewer response drafting

For a broader log format, see the AI tools for research disclosure log. If your work uses interviews or open responses, pair this article with the guide on [[AI disclosure](/how-to-disclose-ai-use-for-neurips-icml-and-acl-submissions/) for qualitative research](/ai-disclosure-for-qualitative-research/).

Participant data changes the rules

Do not paste participant data into a general AI tool unless your ethics approval, consent language, data management plan, and institutional rules allow it.

APA warns authors that organizations running generative AI tools may have access to information entered into them, and that authors should consider how this affects participant privacy, author privacy, and intellectual property. APA also says journal editors and reviewers may not enter submitted manuscript material into generative AI tools because that would violate peer review confidentiality. (apa.org)

Psychology researchers handle data that often looks harmless until combined with context. A single therapy transcript excerpt, classroom observation note, free response, or demographic cluster can identify someone.

Before using AI with participant material, ask four plain questions.

Did the participant consent to this kind of processing?

Did the IRB or ethics committee approve it?

Does the tool provider store or train on the input?

Can you remove or mask identifiers without damaging the analysis?

If you cannot answer those questions, do not upload the data. Use synthetic examples, local tools approved by your institution, or non-sensitive excerpts that no longer identify participants.

For education studies, the same caution applies to student work and classroom data. The related guide on AI disclosure for education research gives examples for minors, teachers, and school settings.

Disclose AI-generated stimuli as part of the method

Psychologists often treat stimuli as instruments.

If AI helped create them, readers need to know. A language model that writes moral dilemmas, chatbot profiles, therapy vignettes, persuasion messages, or social media posts may introduce patterns that affect results.

Do not bury that information in the acknowledgments.

Put it in the method section where you describe materials. Name the tool, describe the prompt strategy, explain how researchers screened or edited the output, and say whether participants saw raw or edited AI text.

A short method note may look like this:

\subsection{Materials}
 
We used ChatGPT to draft an initial pool of 40 workplace conflict vignettes
between March 3 and March 7, 2026. The authors prompted the tool to vary
conflict type, role hierarchy, and outcome valence. Two authors reviewed all
drafts, removed 11 vignettes that contained unclear role descriptions, and
edited the remaining vignettes for length. Participants saw only the edited
versions. The final prompts and selected outputs appear in Supplement A.

This tells readers what the model did and what the authors did.

That distinction matters. The model drafted. The researchers selected, edited, and took responsibility.

If your AI use affected images, figures, or visual stimuli, use the guide on AI-generated images in academic papers as a companion.

AI-assisted coding needs an audit trail

Many psychology projects include text: interviews, diaries, free responses, open survey answers, therapy notes, forum posts, or classroom observations.

AI can help sort text. It can suggest codes, group excerpts, propose themes, or flag contradictions. It can also flatten meaning.

For qualitative psychology, disclose the role of AI in relation to human coding. Did AI create the initial codebook? Did it apply human-made codes? Did researchers compare AI suggestions with human coders? Did the team use the output only during memo writing?

State the boundary.

\subsection{Qualitative analysis}
 
The research team used Claude to suggest candidate codes for 25 anonymized
pilot excerpts. The tool did not code the full dataset. Two authors reviewed
the suggestions, rejected codes that inferred participant intent without textual
support, and developed the final codebook through discussion. All final coding
decisions were made by the authors.

If you report interrater agreement, do not treat an AI tool as a human coder unless your design justifies that choice and your journal accepts it. Explain how the tool entered the workflow.

For interviews, fieldnotes, and open-ended responses, readers should be able to see whether AI influenced the interpretation or only helped with clerical work. The distinction belongs in the methods section, not in a private lab notebook.

Manuscript polishing still needs disclosure when policy asks for it

Some researchers feel odd disclosing grammar help.

That feeling makes sense. Spellcheckers and reference managers rarely appear in disclosure statements. Generative AI tools can do more than fix commas, though. They can change tone, sharpen claims, add transitions, and suggest interpretations.

APA says authors may use AI tools for specific purposes such as editing, but they must disclose use under APA's policy. APA also instructs authors to disclose AI use in the most relevant manuscript location, often the introduction or method section. (apa.org)

If the tool only corrected spelling or formatting, your journal may not need a long note. If the tool rewrote paragraphs, drafted the abstract, suggested claims, or produced text that remains in the paper, disclose it.

A concise disclosure can work:

\section*{AI disclosure}
 
The authors used ChatGPT on May 12, 2026, to edit selected paragraphs for
grammar and clarity. The tool did not generate the study design, hypotheses,
data analysis, results, or interpretation. The authors reviewed and approved
all final text.

If you need more examples, see How to disclose ChatGPT usage in academic papers and [AI Usage Cards examples and templates](/ai-usage-cards-examples/).

APA journals, authorship, and responsibility

AI tools do not meet authorship requirements.

APA says AI cannot be named as an author on an APA scholarly publication. COPE also states that AI tools cannot be listed as authors because they cannot take responsibility for submitted work. (apa.org)

This rule helps psychology authors write cleaner statements.

Do not write "ChatGPT contributed to this paper as a co-author." Write what the tool did.

For example:

\section*{Acknowledgments}
 
The authors used Microsoft Copilot to draft R code for descriptive statistics
and used ChatGPT to edit selected sentences for clarity. The authors checked
the code against the study protocol, verified all outputs, and take
responsibility for the final manuscript.

The human authors own the study. They designed it, checked it, interpreted it, and submitted it.

If you need to compare journal rules before submission, start with AI disclosure policies by major journals and AI transparency requirements for journal submissions.

Where to put the disclosure in a psychology paper

The best location depends on what AI changed.

If AI helped design stimuli, place the disclosure in the method section under materials.

If AI helped analyze data, place it under analysis.

If AI helped write or edit the manuscript, use an AI disclosure statement, acknowledgments section, or the location required by the journal.

If AI helped during peer review, follow the journal's reviewer policy. APA bars reviewers and editors from entering submitted manuscript materials into generative AI tools because peer review materials are confidential. (apa.org)

The CDC's 2026 guidance also frames disclosure across the scientific workflow, including proposing, preparing, performing, and reviewing scientific products. That lifecycle view fits psychology well because AI use often starts before the first draft. (cdc.gov)

A practical rule works: place the disclosure next to the research step that AI affected.

Readers should not have to hunt.

What to include in an AI Usage Card for psychology

An AI Usage Card gives you a compact record before you write the final disclosure.

For a psychology paper, include the study stage, tool, model, date, input type, output type, human review, privacy controls, and manuscript location. If the study involved participants, describe whether you entered participant data into the tool. If you did, name the approval basis and safeguards.

A card for a survey experiment might say:

\section*{AI Usage Card}
 
\textbf{Tool:} ChatGPT, accessed through an institutional account.
 
\textbf{Date of use:} February 10 to February 18, 2026.
 
\textbf{Purpose:} Drafted candidate workplace vignettes for a survey experiment.
 
\textbf{Inputs:} Researcher-written prompts describing vignette constraints.
No participant data were entered.
 
\textbf{Outputs used:} Initial vignette drafts. The authors edited all final
stimuli for length, readability, and construct fit.
 
\textbf{Human review:} Two authors screened all drafts and excluded items that
introduced unintended moral judgment or unclear role hierarchy.
 
\textbf{Manuscript location:} Materials subsection and Supplement A.

You can generate this structure at ai-cards.org and adapt the text for your manuscript, supplement, preregistration, or lab archive.

If your paper uses LaTeX, the LaTeX AI Usage Cards tutorial and Overleaf guide show how to add the card without breaking your template.

Common mistakes in psychology AI disclosures

The first mistake is writing "AI was used" and stopping there.

That tells readers almost nothing. Name the tool and the task.

The second mistake is hiding stimulus generation. If participants responded to AI-generated text, images, scenarios, or chatbot messages, that belongs in the method.

The third mistake is uploading sensitive data before checking consent and ethics approval. Psychology data can identify people through context, even after you remove names.

The fourth mistake is treating AI output as evidence. A model can help draft code or summarize text. It does not verify a theory, replace participant data, or settle an interpretation.

The fifth mistake is forgetting the response letter. If AI helped draft replies to reviewers, record that use too. The same accountability logic applies.

For broader ethical framing, see AI ethics and documentation in academic research and Why AI transparency matters in research.

A short checklist before submission

Before you submit a psychology manuscript, check five things.

Did AI affect the design, materials, analysis, writing, figures, code, or responses to reviewers?

Did you record the tool name, version if known, date, prompts, and output type?

Did any input include participant data, clinical material, student work, or unpublished peer review material?

Did your disclosure appear in the manuscript section where readers need it?

Did every human author review and accept responsibility for the final text, analyses, and claims?

If you can answer those questions, your disclosure will read like part of the research method rather than an apology tacked on at the end.

Generate an AI Usage Card before you submit. Use it as a supplement, paste its text into your AI disclosure statement, or keep it in your project folder so you can answer editor questions without reconstructing your AI use from memory.