Understanding what each is and when to use it.

Citation (references in your bibliography):

• When to use: When you directly quote an AI system's output or substantially use information from its output
• Example: If you ask ChatGPT "What are the main theories of X?" and use its response as a theory discussion in your paper
• Format: Cite as [Tool Name] or the company (varies by style guide)
• Purpose: Allows readers to find the source material

Attribution (acknowledgment in text):

• When to use: When you want readers to know that an AI tool contributed to your work
• Example: "We used Claude to help organize the literature into themes" or "GitHub Copilot assisted in code generation"
• Format: In-text note or acknowledgments section
• Purpose: Clear statement of who/what contributed to your work

Transparency statement (usually in methods or acknowledgments):

• When to use: For substantive AI use affecting research methods or conclusions
• Example: "AI-assisted literature synthesis using Semantic Scholar and Claude to organize papers by methodology; the authors verified all synthesis and identified synthesis limitations"
• Format: Usually 1-3 sentences describing what AI did and how it was reviewed
• Purpose: Comprehensive disclosure allowing readers to assess rigor

What's required where:

• Coursework: Check syllabus/instructor preference
• Journal publication: Check journal's author guidelines
• Dissertations: Check institutional guidelines and advisor preference
• Grants/proposals: Check funder guidelines
• When in doubt: Full transparency is safer than partial

How major style guides address AI citation (standards are evolving; verify current versions).

APA Style (American Psychological Association):

• Cite as: Author (as company name), Year (e.g., OpenAI, 2024; Anthropic, 2024)
• Format: Company Name. (Year). Tool name (Version if available) [Large language model].
• Example in text: "According to OpenAI (2024), ChatGPT can be used for...analysis."
• Example in reference list: "OpenAI. (2024). ChatGPT (Version 4) [Large language model]. https://www.openai.com/chatgpt"
• Best practice: Include retrieval date and URL

MLA Style (Modern Language Association):

• Cite as: Company Name or Tool Name (MLA 9 is still developing standards)
• Format: Tool Name, Company Name, Year, website/tool
• Example: "Claude, Anthropic, 2024, www.claude.ai"
• Note: MLA standards are evolving; check current version

Chicago Style (Chicago Manual of Style):

• Approaches vary; typically cite tool as corporate author
• Format: Tool Name, Company, Access date
• Example: "ChatGPT, OpenAI, accessed January 15, 2024, https://chatgpt.openai.com"
• Note: May go in text, footnote, or bibliography depending on use

Journal-specific guidance:

• Nature: "Disclose in methods if AI contributed substantially"
• Science: "Any use of AI or algorithms must be disclosed"
• JAMA: "Authors should disclose use of generative AI in manuscript preparation"
• American Journal of Psychiatry: "Any use of artificial intelligence or machine learning should be explicitly stated"
• Verify your journal's specific requirements

How to write clear, accurate transparency statements.

Core elements:

1. What tools were used: Specific names and versions if available
2. What they did: Specific tasks or sections where tools contributed
3. How output was reviewed: What verification or oversight happened
4. Limitations acknowledged: What the AI couldn't verify or limitations of its contributions

Example statements:

Light AI use (minor contributions):

• "Grammarly was used for grammar and style checking of the final manuscript."

Moderate AI use (more substantial):

• "Semantic Scholar and Elicit were used to identify relevant papers in the initial literature search. ChatGPT was used to help organize papers by methodology. All papers and organizational decisions were verified by the authors."

Heavy AI use (significant contribution to multiple areas):

• "This research used AI tools for three primary purposes: (1) Literature discovery using Semantic Scholar and Elicit to identify 150 initial papers; authors screened all papers manually for relevance. (2) Code generation using GitHub Copilot for analysis; all code was reviewed for correctness and statistical appropriateness by [Statistician name]. (3) Writing improvement using Claude for clarity; all scientific content was verified by the authors. AI tools did not participate in hypothesis generation, research design, or results interpretation, which remained under full author control."

Best practice principles:

• Be specific about what tools did
• Acknowledge what humans verified
• Be honest about limitations
• Avoid overstating AI's role (it's a tool, not an author)
• When in doubt, be more transparent rather than less

Understanding that AI tools are not authors and you remain responsible.

Key principle: AI tools are not co-authors, regardless of how much they contributed.

Why:

• Tools cannot make intellectual decisions (what matters, what's important)
• Tools cannot bear responsibility for accuracy or ethics
• Tools cannot defend work or respond to criticism
• Authorship requires intellectual responsibility, which AI doesn't have

What this means:

• You (human) are responsible for all content, including AI-generated text
• You must verify accuracy, appropriateness, and ethics
• You must disclose AI contributions but remain sole author
• You cannot delegate responsibility to the tool

Acknowledgments vs. Authorship:

• Acknowledgments: List people/tools that helped but didn't make intellectual contributions
• Authorship: For people who made intellectual contributions, not tools
• Example: "We thank Claude for assistance with literature organization. [Human contributors] designed the study and interpreted results."

How to document AI use so others can understand and reproduce your work.

What to document:

• Tool and version: "Claude 3 Sonnet, accessed January 2024"
• Model parameters used: "Temperature set to 0.5 for consistent data extraction"
• Prompts used: The actual text you sent to the AI (especially if tool use was substantive)
• Outputs reviewed: What proportion of outputs you reviewed? What was your rejection rate?
• Verification process: How did you check AI outputs for accuracy?
• Limitations: What could the AI not verify? What required your expertise?

Where to document:

• Methods section: Describe AI tool use as part of your methodology
• Supplementary materials: Provide example prompts, example outputs, documentation of review process
• GitHub/OSF: Many researchers post their actual prompts and outputs for full transparency
• Appendices: Full documentation of AI use workflow

Example methods section language:

"We used Claude (Anthropic, version 3.5 Sonnet) to assist in literature organization. We conducted a comprehensive literature search using Semantic Scholar and submitted all 200 identified paper titles and abstracts to Claude with the following prompt: [exact prompt]. We reviewed all AI responses for accuracy, categorizing suggestions using [your criteria]. We accepted 85% of AI categorizations without modification; remaining 15% were discarded due to incorrect thematic assignment or duplicate categorization. All final categorization was reviewed by [Team member] and [Team member] for consensus."

Practical Research Use Cases

Scenario: You've written a paper where:

• Claude helped organize your literature synthesis
• GitHub Copilot generated analysis code (which you reviewed)
• Grammarly improved grammar
• ChatGPT provided general writing advice

Disclosure approach:

In Methods section:

"We used the following AI tools in our research: (1) Claude was used to organize 45 selected papers into thematic categories based on methodology; all categorizations were verified by [Lead author]. (2) GitHub Copilot was used to generate initial code for mixed-effects analysis; all code was reviewed for statistical correctness by [Statistician] and validated on a subset of data. (3) Grammarly was used for grammar and style checking of the final manuscript."

In Acknowledgments:

"We acknowledge the use of Claude, GitHub Copilot, and Grammarly in preparing this manuscript. All substantive content was generated by the authors."

In References (if used to generate specific content):

[Include citations if any AI-generated output was substantially used and cited]

Verification of journal requirements:

• Check target journal's policy
• Format disclosure statement per journal requirements
• If journal requires specific format, follow it

Scenario: Your dissertation used:

• AI for literature search (major time-saver)
• AI for code generation
• AI for writing improvement
• AI tools not mentioned to advisor

Disclosure approach (note: should have discussed with advisor first):

Chapter 1 opening or methods chapter:

"This dissertation integrated artificial intelligence tools into the research process for efficiency while maintaining full researcher responsibility and verification. Specifically:

Literature Discovery: Semantic Scholar and Elicit were used to identify papers related to [research question]. Initial searches returned 500+ papers; the author manually screened all titles and abstracts for relevance, ultimately selecting 50 papers for full-text review. This AI-assisted process reduced literature screening time from an estimated 40 hours (traditional approach) to 12 hours while increasing comprehensiveness.

Data Analysis: GitHub Copilot assisted in generating code for statistical analysis. All generated code was reviewed for correctness by [Committee member], a biostatistician, who verified statistical assumptions and appropriateness of chosen tests.

Writing: Claude was used to improve clarity of written sections. The author reviewed all suggested changes and only accepted modifications that improved clarity without changing scientific content.

All research design decisions, hypothesis generation, interpretation of results, and conclusions represent the author's original thinking."

Scenario: You used AI to help write a grant proposal and want to disclose this.

Disclosure approach:

Include in proposal (check specific funding agency requirements):

"The applicant used ChatGPT to assist with initial drafting of specific sections of this proposal. All scientific content, proposed methods, and intellectual merit were generated by the applicant and reviewed for accuracy and consistency. AI tools did not participate in hypothesis generation, methodology design, or evaluation of feasibility."

Or in a separate footnote:

"Note on tool use: This proposal was prepared with assistance from [tools]; all scientific content was verified by the applicant."

Best practice: Call program officer before submitting if unsure about requirements.

Scenario: You're presenting research at a conference and want to transparently discuss AI use.

Disclosure approach:

Include in methods slide:

"AI Tool Use in This Research:

• Literature search: Semantic Scholar (200+ initial papers)
• Code generation: GitHub Copilot (reviewed by [statistician])
• Writing improvement: Claude (clarity enhancement)
• All research design and interpretation: Author"

Or in final Q&A slide:

"Transparency Note: This research used [specific tools] for efficiency. All scientific decisions, verification, and responsibility remain with the author."

Benefit: Transparency prevents audience suspicion and demonstrates sophisticated research practice.

Hands-On Exercise

Objective: Document your own AI use in a way that's clear, transparent, and reproducible.

Steps (Time: 45-60 minutes):

1. Audit your AI use (15 minutes):

• For your current research project, list every AI tool you've used
• For each tool: What did you use it for? How much did it help?
• Document: Tool name, version, specific tasks, role in your research

1. Create a transparency statement (15 minutes):

• Write a 2-3 sentence statement disclosing your AI use
• Include: What tools, for what tasks, how output was verified
• Make it honest and specific without overstating AI's role
• Avoid: "AI was used" (too vague); Include: Specific tools and tasks

1. Check style guide requirements (10 minutes):

• Identify: Will you submit to journal? Which one?
• Check: Journal's AI policy/requirements
• Check: Required citation format (APA, MLA, Chicago, other)
• Document: How to cite your AI tools in your field's format

1. Create full documentation (15 minutes):

• Write a paragraph for your methods section describing AI use
• Include: What tools, what they did, how you reviewed output, limitations
• Make it something you could include in a published paper
• Include: Enough detail that someone could understand your workflow

1. Get feedback (Optional, 10-15 minutes):

• Share your documentation with your advisor
• Ask: "Is this clear? Do I need to disclose more? Is this format appropriate for our field?"
• Incorporate feedback

Time required: 45-60 minutes

Common Mistakes and Misconceptions

Current standards are moving toward treating tools as citable tools, not published works. When you use an AI tool substantively, acknowledging it (even if not formally citing) is increasingly expected.

Reviewers often ask about AI use or notice its fingerprints in writing/code. Proactively disclosing is better than being caught hiding it.

It won't. Most journals now expect AI disclosure. Failure to disclose when required can result in rejection or retraction. Transparency is the safe approach.

Absence of prohibition isn't permission. Good practice: ask before using AI substantially. This prevents surprises later.

No. Authorship requires intellectual responsibility. AI tools cannot defend work, make decisions about what matters, or bear responsibility for accuracy. Acknowledge substantial contributions but don't list AI as author.

Key Takeaways

• Citation and attribution standards are evolving: APA, MLA, and Chicago have guidance; journals have developing policies; verify requirements for your specific context
• Transparency statements disclose what AI did, how output was verified, and limitations: These are increasingly expected in research publications
• AI tools are tools, not authors: Acknowledge contributions but maintain that humans are responsible for all content; AI is cited or acknowledged, not listed as author
• Documentation enables reproducibility: Recording exact tools, versions, prompts, review processes, and limitations allows others to understand and potentially replicate your work
• Specificity matters: Say "Claude was used to organize papers by methodology; all categorizations were verified by [author]" rather than vague "AI was used"
• Different contexts have different requirements: Check journal, institution, and funder guidelines; when in doubt, be more transparent rather than less

Reflection Questions

1. Your publication plans: Where do you plan to publish? What does that journal/venue require about AI disclosure?

1. Your AI use documentation: For your current research, what would you need to document to fully disclose your AI use? What are the key details readers need to know?

1. Citation approach: For your field's style guide (APA, MLA, Chicago, other), how would you cite an AI tool you used substantially? Practice writing the citation.

1. Transparency statement: Write a 2-3 sentence transparency statement for your research that you could include in a paper, dissertation, or presentation.