Two different questions get asked as if they were one. Declaring AI use means telling your reader that you used a tool, which tool, and for what. Citing AI means putting an entry in the reference list. The declaration is the one that matters in a Philippine thesis, the one your school’s integrity rules are actually about, and the one to get right first. The reference entry is a formatting question that follows from it.
The honest state of APA’s guidance
This deserves stating plainly rather than glossed over. As of this writing, APA has not issued a full guideline in the Publication Manual for generative AI; what exists is a blog post about ChatGPT. The most detailed treatment available is a supplement published by the Werkgroep APA through SURF, De APA-richtlijnen uitgelegd: Generatieve AI (version 1.0), which builds on the recommendations in that blog post and extends them to inputs and outputs that are not text. That supplement is the source of the reference structure set out below, and it says in its own introduction that an official APA guideline is still outstanding.
Two things follow. First, anyone who tells you there is one settled APA rule for citing ChatGPT is overstating it. Second, and more usefully, the reasoning behind the guidance is stable even while the formatting is not: AI output is not retrievable by anyone else, and the same prompt produces different output on different runs, so it cannot be treated as a source a reader could go and check.
Why that reasoning changes what you do
A reference entry exists so a reader can retrieve what you read. AI output fails that test at its root. The practical consequence, and the thing to understand before any formatting question, is this: generative AI is described as part of your method, not relied on as a source of knowledge. If a factual claim in your thesis rests only on something a chatbot said, the claim is unsupported, and the fix is not a better citation but a real source. The referencing problem people ask about is really a sourcing problem wearing a disguise.
That is the principle. Where it lands in your own manuscript is decided by your school.
Your school’s rule binds, and you have to go and read it
There is no single national Philippine rule on AI use in student work that supersedes institutional policy. Universities are setting their own. The University of the Philippines, for instance, has adopted a set of Principles for Responsible and Trustworthy Artificial Intelligence covering the common good, empowerment, cultural sensitivity, transparency, accountability, fairness and safety, together with an AI Advisory Committee and an AI Advisory Board. It is an institutional governance document rather than a student citation rule, and it does not by itself tell you how to write a declaration in Chapter 3. Other universities have issued guidance at the college or department level, which is often where the binding student-facing rule actually lives.
So, before you write anything: get your own rule in writing, from your research coordinator or your college. Three questions answer everything else.
- Is AI use permitted at all, and for what? Some departments permit language editing and prohibit drafting. Some permit neither. Some are silent, which is not the same as permission.
- Is a declaration required, and in what form? Many schools now require a signed statement or a paragraph in the manuscript. Some prescribe the wording.
- Where does it go? The certification of originality, a separate AI-use declaration page, the methodology, or an appendix.
A rule you were told verbally is worth less than a rule you can produce at a defence. Ask for the memo.
Where the disclosure sentence goes

The APA supplement draws a line that maps neatly onto how students actually use these tools, and the line decides where the disclosure lives.
| How you used it | Where you say so | What you must include |
|---|---|---|
| As a writing aid: checking spelling, adjusting phrasing, simplifying a sentence | The introduction, or your school’s declaration page | The tool and what it did. The prompts do not need to be reproduced |
| As a research tool: generating question wording, suggesting structure, processing material | The methodology section | The tool, a description of the input, and a description of the output |
Where the tool was a research tool, the supplement asks for the prompt to be recorded, in an appendix if it is long or if there are several, and for the output to be recorded too, wholly or in the relevant part. Where a tool offers a shareable link to the conversation, that link may be given in the reference list instead of reproducing a copy. It also suggests stating how often and over what period the tool was used, and recording any settings you changed, at the level of detail your programme asks for.
Writing-aid disclosure, written out.
Language editing support was used in preparing this manuscript. [Tool name] ([Developer], [year]) was used to check spelling and to adjust sentence phrasing in Chapters 1 to 5. All content, analysis and conclusions are the researchers’ own.
Research-tool disclosure, written out.
In refining the wording of the research questions, [tool name] ([Developer], [year]) was used. The prompt given was: “[prompt]”. The output returned is reproduced in Appendix [letter]. The output was used only to compare phrasings; the final research questions were formulated by the researchers and approved by the adviser. The tool was used on [number] occasions during [period].
Notice what both paragraphs do. They name the tool, they say exactly what it touched, and they draw a boundary around it. That boundary is what a panel is actually checking.
The reference entry, when one is needed
The supplement gives the entry six elements, in this order.
| # | Element | Note |
|---|---|---|
| 1 | Developer | The company that provides the tool |
| 2 | Year | Use n.d. where there is none |
| 3 | Name of the tool | In italics |
| 4 | Version | In round brackets; omit where no version number is given |
| 5 | Description | In square brackets: [Generative AI] |
| 6 | URL | No retrieval date |
Pattern. Developer. (Year). Name of tool (Version) [Generative AI]. URL
Example, as the supplement gives it. OpenAI. (2024). ChatGPT (Version 3.5) [Generative AI]. https://chat.openai.com
In text. Parenthetical: (OpenAI, 2024). Narrative: OpenAI (2024). Note that the in-text citation carries the developer, not the tool name, which surprises most people the first time. The tool name is italicised on its first mention in the running text.
Where the tool has no stated year, the entry takes n.d. in the year position. Handling a source with no date is a general APA problem, not an AI-specific one, and the rule is the same here as everywhere else.
Images and other non-text output
If a generative tool produced a figure in your manuscript, the supplement’s rule is that the note beneath the figure names the prompt, the developer and the tool, with the tool’s URL, and the tool gets a reference list entry. It also notes that no copyright statement is placed under an AI-generated image, because ownership of AI-generated output is unsettled. For a Philippine thesis this is mostly an edge case, but it is worth knowing before you drop a generated diagram into Chapter 2. The research paradigm, incidentally, should not be one; it is a figure you construct from your own variables, as the paradigm examples show.
What a panel will actually ask
- “Did you use AI, and for what?” Answer from your declaration, not from memory. This is why the declaration is written before the defence.
- “Where did this claim come from?” Every factual claim traces to a real source. If one traces only to a chatbot, remove it or find the source. This is the question that ends badly for students who used AI as a library.
- “Are these references real?” Generative tools produce plausible references to work that does not exist. Open every reference you cite. A fabricated reference is the most serious error on this list, and it is the one most likely to be discovered.
- “Did you upload data or the manuscript to the tool?” Respondent data carries obligations under the Data Privacy Act of 2012, Republic Act No. 10173, and uploading identifiable data to a third-party service is a decision your ethics arrangements have to cover.
That last point is the one most often missed. Consent obtained for a study does not automatically extend to processing the data through an external service. Where that matters most, and how consent is documented, is set out in the piece on ethics approval and consent, and the citation mechanics for the statute itself are in the piece on citing Republic Acts and Supreme Court decisions.
What the tools can and cannot do for a thesis
Separately from the citation question, it is worth being clear-eyed about capability. The honest assessment of what an AI thesis tool does well and what it cannot supply is set out in the piece on AI thesis makers in the Philippines, and the comparison across the available tools is in the best AI tools for thesis writing. If your concern is the similarity report rather than the citation, what a pre-check can and cannot see is covered in the piece on plagiarism checkers compared. And the reference list the tool helped you assemble still has to survive a format check, which is what the guide to APA 7 format for a Philippine thesis covers.
Write the declaration before you need it
The declaration is four sentences, and writing it early forces the useful question: what exactly did the tool do, and does every claim in my thesis still have a real source behind it? Tesify is built for that answer to be yes. It drafts from your outline, your data and the sources you supply rather than inventing them, keeps every reference attached to something you added, and leaves the research and the conclusions entirely yours, which is what makes a declaration straightforward to write.
Draft your chapter free on Tesify
Frequently asked questions
Does APA 7 have an official rule for citing ChatGPT?
Not a full one. APA has published a blog post on ChatGPT rather than a complete guideline in the Publication Manual. The most detailed available treatment is the Werkgroep APA supplement on generative AI published through SURF, which builds on that blog post and states that an official APA guideline is still outstanding.
Is using AI for a thesis allowed in the Philippines?
There is no single national rule that supersedes institutional policy; each university and often each college sets its own. Get your department’s rule in writing before relying on anything.
Do I put ChatGPT in my reference list?
Where your school permits AI use and you used a tool in a way that needs recording, yes, following the six-element pattern. But the declaration describing what the tool did is the more important element, and a factual claim should never rest on AI output alone.
What does the reference entry look like?
Developer. (Year). Name of tool (Version) [Generative AI]. URL. For example: OpenAI. (2024). ChatGPT (Version 3.5) [Generative AI]. https://chat.openai.com
What goes in the in-text citation?
The developer and the year, as (OpenAI, 2024) or narratively as OpenAI (2024). Not the tool name, which is italicised only on its first mention in the running text.
Where do I put the prompts?
In an appendix where they are long or numerous, or in the text where a single short prompt is relevant. Where the tool provides a shareable link to the conversation, that link may go in the reference list instead.
Do I have to declare it if I only used it to fix grammar?
The APA supplement treats a writing aid as something you mention in the introduction without reproducing prompts. Whether your school requires a formal declaration for language editing is a departmental question; ask.
Will my adviser know I used AI?
Detection tools are unreliable in both directions, which is exactly why declaring is the safer course. A declared use is a methodological choice; an undeclared one that surfaces later is an integrity finding.
Can I cite an AI-generated summary of a study?
No. Cite the study, which means opening it. A summary you did not verify may describe a paper that does not exist, and fabricated references are the most serious error in this area.
Is it safe to upload my data to an AI tool?
Treat it as a data protection decision, not a convenience one. Respondent data carries obligations under the Data Privacy Act of 2012, Republic Act No. 10173, and consent to participate in a study does not automatically extend to processing that data through an external service.
