AI prompts for desk research: search, extraction, synthesis, verification

These prompts help product and UX teams use AI to run desk research faster without losing source quality. Replace [bracketed placeholders] with your specifics before pasting into ChatGPT, Claude, Perplexity Deep Research, or another LLM.

Prompt 1: Run a structured desk research search

I am running desk research for a [UX / product / market / service] project on [topic].

Research question: [the specific question the desk research should answer]
Audience: [who the findings are for and what decision they will inform]
Scope constraints: [geography, time period, source types to include or exclude]

Please:
1. Identify 5-8 subtopics that need to be covered to answer the research question fully
2. For each subtopic, recommend 2-3 source types to search (academic databases, industry reports, competitor blogs, government data, online communities, etc.)
3. Suggest specific search queries for each subtopic that would surface credible sources
4. List 5-10 well-known publications, research firms, or databases that are likely to have relevant material on this topic
5. Flag any topics where the literature is likely to be sparse or unreliable, so I can plan to cover them through primary research instead

Prompt 2: Extract structured findings from a source

Here is a source I am including in a desk research project:

Title: [title]
Author: [author]
Publication: [where it appeared]
Date: [date]
Full text or URL: [paste or link]

My research question: [the specific question]

Please extract:
1. The main claims relevant to my research question (quote directly, do not paraphrase)
2. The sample size, geography, and method behind each claim
3. The funding source or institutional affiliation of the author (potential bias)
4. Any limitations the author flags about their own findings
5. The single most important finding from this source as it relates to my question
6. A credibility score from 1-5 with one-sentence justification
7. Two follow-up questions this source raises that another source might answer

Prompt 3: Synthesize across multiple sources

I have extracted findings from [N] sources for a desk research project on [topic]. The research question is [specific question].

Here are the structured findings:
[Paste the extracted findings from each source, with source name attached to each finding]

Please:
1. Group the findings by subtopic and identify the dominant pattern in each subtopic
2. Highlight where multiple sources agree (consensus findings)
3. Highlight where sources contradict each other and suggest why (different sample, geography, time period, method)
4. Identify gaps where no source covers a relevant question — these become the input for primary research
5. Flag any finding that appears in only one source and is not corroborated — treat with caution
6. Draft a one-paragraph executive summary that ties the findings to the research question
7. Recommend the top 3-5 questions the team should validate through primary research

Prompt 4: Verify citations and check for hallucinations

I am about to publish a desk research brief that cites [N] sources. Before I send it out, I need to verify every claim.

Here is the brief:
[Paste the brief]

For each cited claim:
1. Identify the exact source it is attributed to
2. Flag any claim where the citation looks suspicious (vague title, missing date, unfamiliar publication)
3. Suggest a verification step for each claim — which page of the source to check, which dataset, which URL
4. Identify any claim that does not have a clear source attached and would need to be removed or sourced
5. Suggest 3-5 places where the brief could be strengthened by adding a corroborating source
6. List any claims that contradict other findings in the brief but are not flagged as contradictory