How to disclose Perplexity AI use in academic research

Perplexity changes the disclosure question

Perplexity feels less like a chatbot and more like a search tool.

That makes disclosure easy to forget.

A researcher may ask Perplexity for recent papers on a method, compare definitions across sources, summarize an uploaded PDF, or draft a short paragraph based on web results. None of that looks dramatic in the moment. It feels like search with a better interface.

But the research record still needs a trace.

Perplexity describes its service as an AI powered search engine and says users can submit input to the Perplexity Engine, which then uses AI tools to generate output based on that input. Its own product page also presents Perplexity as a source-citing answer engine. That mix of search, synthesis, and generation matters for disclosure. (perplexity.ai)

If you used Perplexity during research or writing, record what you asked it to do, what sources you checked, and how its output affected the final work.

Then generate an [[AI Usage Card](/ai-disclosure-for-social-science-research/)](/) and attach it to your manuscript files, thesis appendix, preregistration materials, lab notebook, or submission cover letter.

When Perplexity use needs disclosure

Disclose Perplexity when it shaped the scholarly work.

That includes literature discovery, source comparison, paper summaries, argument mapping, code explanation, data interpretation support, or text drafting. If Perplexity helped you decide what to read, what to cite, how to frame a claim, or how to phrase a paragraph, readers deserve to know.

You can treat very small uses differently. If you used Perplexity once to check the spelling of a software package name and then verified it elsewhere, a manuscript disclosure may not add much. Still, a private log helps. Small uses become hard to reconstruct after a paper has gone through revisions.

For a broader decision guide, see Do I need to disclose AI usage in my paper?. If you already know you need a statement, compare this page with How to disclose ChatGPT usage in academic papers. The logic overlaps, but Perplexity adds one extra issue: cited search results can create a false sense of certainty.

A citation in an AI answer is a lead. It is not proof that the answer represents the source.

Open the source. Read the relevant passage. Cite the original paper, dataset, or policy, not the Perplexity answer.

What to record while you work

The best disclosure starts before submission.

Open a plain text file, spreadsheet, or project note when you first use Perplexity. You do not need a perfect diary. You need enough detail that another researcher can understand the role of the tool.

Record the tool name, date, account type if relevant, mode or model if shown, prompts or short prompt summaries, uploaded files, and what you did with the output. If Perplexity returned sources, record which sources you opened and verified.

A practical entry might look like this:

Date: 2026-10-05
Tool: Perplexity AI
Task: Search for recent papers on synthetic control methods in education policy.
Input: Query about post-2022 applications and limitations.
Output used: Helped identify three candidate papers for screening.
Verification: Opened publisher pages and checked abstracts, methods sections, and DOIs manually.
Final manuscript impact: No generated text copied. Two verified papers entered the literature review.

This kind of record pairs well with the disclosure log in AI tools for research. It also gives you clean material for an AI Usage Card later.

Do not store sensitive data in the log unless your institution allows it. For human subjects research, describe the category of information instead of pasting raw prompts that contain participant details.

How to write a Perplexity disclosure statement

A good disclosure statement names the tool and tells the reader what the tool did.

It also tells the reader what you did after the tool responded.

Weak disclosure:

The authors used AI during the preparation of this manuscript.

Better disclosure:

The authors used Perplexity AI on 5 October 2026 to identify candidate literature on synthetic control methods in education policy and to summarize search results. The authors opened and checked the cited sources before using them. No Perplexity-generated text appears in the manuscript.

For a stronger record, add the AI Usage Card as a supplement or appendix. The statement can stay short because the card carries the workflow details.

If your journal asks for a declaration section, use its required wording first. Publisher policies differ. Springer Nature, for example, asks authors to describe AI use and confirm author accountability in its manuscript preparation guidance. COPE states that AI tools cannot qualify as authors because they cannot take responsibility for the submitted work. (springernature.com)

That last point stays simple: Perplexity can appear in a disclosure statement. It cannot appear in the author list.

Example disclosure statements for common Perplexity uses

Literature search

Use this when Perplexity helped you find papers, books, policies, datasets, or reports.

The author used Perplexity AI in October 2026 to identify candidate sources on [topic]. The author treated Perplexity outputs as search leads, opened the cited sources, checked bibliographic details, and cited only sources that were reviewed directly.

This statement works well when the tool influenced the search process but did not write the manuscript.

If your project involves a structured evidence search, take more care. For systematic reviews, scoping reviews, and meta-analyses, AI assisted search can affect inclusion decisions. Pair this page with [[AI disclosure](/how-to-disclose-ai-use-for-neurips-icml-and-acl-submissions/) in systematic reviews and meta-analyses](/ai-disclosure-in-systematic-reviews-and-meta-analyses/) before you submit.

Source summaries

Use this when Perplexity summarized papers, webpages, policies, or reports.

The author used Perplexity AI to generate preliminary summaries of selected sources. The author checked the summaries against the original sources and wrote the final synthesis without copying Perplexity output.

This tells the reader that the tool helped with reading support, not evidence verification.

If you copied or adapted any generated phrasing, say so.

The author used Perplexity AI to draft short summary notes for selected sources. The author revised those notes after checking the original sources. Portions of the literature review draw on these revised notes.

That version feels less tidy. It is also more honest when the workflow blended summary and drafting.

Manuscript wording

Use this when Perplexity helped with wording, structure, or short passages.

The authors used Perplexity AI to suggest alternative wording for selected paragraphs in the introduction. The authors reviewed, edited, and approved all final text. The tool did not generate results, interpretations, or conclusions.

This kind of statement resembles disclosure for other writing assistants. See [How to disclose Claude use in academic writing](/how-to-disclose-claude-use-in-academic-writing/) and How to disclose Google Gemini use in academic writing if your project used more than one system.

Methods or code help

Use this when Perplexity explained code, suggested packages, or helped compare methods.

The authors used Perplexity AI to locate documentation and examples for [software package or method]. The authors checked the linked documentation and tested all code before use. Perplexity did not analyze the study data.

If Perplexity helped write code that affected results, add that detail. Researchers should know whether the tool touched analysis logic, data cleaning steps, or visualization code.

For computer science and computational work, see AI disclosure for computer science research papers.

How to include an AI Usage Card in LaTeX

After you generate a card at ai-cards.org, you can include it as a figure, an appendix item, or a supplemental file.

For a short paper, use a compact appendix:

\appendix
 
\section{AI usage disclosure}
 
The authors used Perplexity AI on 5 October 2026 to identify candidate
sources and summarize search results during the literature review.
The authors checked all cited sources against the original publications
before citation. No AI-generated text was copied into the manuscript.
 
\begin{figure}[h]
  \centering
  \includegraphics[width=0.95\linewidth]{ai-usage-card.pdf}
  \caption{AI Usage Card for the manuscript preparation workflow.}
  \label{fig:ai-usage-card}
\end{figure}

If the journal discourages figures in appendices, keep the text statement in the manuscript and upload the card as a supplement.

For more formatting options, use the LaTeX Tutorial for AI Usage Cards or the Overleaf guide.

What an AI Usage Card adds

A disclosure statement answers the journal's immediate question.

An AI Usage Card answers the reader's next question.

What did you ask the tool to do? Did it touch the data? Did it draft text? Did it help select sources? Did you verify the output? Did a human author approve the final material?

Those details rarely fit in a short declaration. They do fit in a card.

An AI Usage Card also helps coauthors. One author may use Perplexity for literature search, another may use a different tool for code comments, and a third may use no AI at all. The card gives the team one place to record the shared workflow before submission pressure makes everyone guess from memory.

If you want examples before you make one, see AI Usage Cards examples and templates and What are AI Usage Cards?.

Privacy and source integrity checks

Perplexity can process prompts, uploads, and outputs. Its general terms define input and output for the Perplexity Engine, while its API documentation states a zero data retention policy for the Chat Completions API. Enterprise help pages describe separate retention controls for organizations. These differences matter because a personal account, API use, and enterprise deployment may follow different terms. (perplexity.ai)

Before you paste anything into Perplexity, ask two questions.

First, does the prompt contain confidential data, unpublished findings, peer review material, student work, patient data, interview excerpts, or proprietary information?

Second, do your university, funder, ethics board, or journal rules allow that use?

If the answer feels uncertain, do not paste the material. Rewrite the prompt at a higher level or use an approved institutional tool.

Source integrity needs the same care. Perplexity may return useful links, but you should verify bibliographic details, authorship, publication status, DOI, page numbers, and the claim you plan to cite. If a source matters to your argument, read it yourself.

A Perplexity disclosure workflow you can finish today

Start with your actual uses. Do not write a grand statement about AI. Write a record.

List each Perplexity task in plain language:

1. Found candidate sources for the literature review.
2. Summarized search results for screening.
3. Suggested alternative wording for two introduction paragraphs.
4. Helped locate package documentation for the analysis script.

Then mark each task as "used in final work" or "discarded."

For every "used" item, add your verification step:

1. Candidate sources: checked original papers and DOIs.
2. Search summaries: compared against source abstracts and methods sections.
3. Wording suggestions: rewritten by authors and checked for meaning.
4. Package documentation: tested code and cited official documentation.

Now turn that log into two outputs.

Write a short manuscript statement.

Generate an AI Usage Card at ai-cards.org.

You can copy the card text into a cover letter, add the PDF to supplementary materials, or keep it in your project archive in case an editor asks for more detail. If your paper goes through several revisions, update the card before resubmission.

The cleanest rule

Disclose Perplexity when it shaped the work.

Do not cite Perplexity as evidence. Cite the sources you checked.

Do not name Perplexity as an author. Name it as a tool.

And do not wait until the submission portal asks for an AI statement. By then, the details blur.

Generate an AI Usage Card now, while you still remember what you asked, what you checked, and what changed in the manuscript.