How to Use AI for Research

AI can cut your research time in half. But used poorly, it can also fill your work with errors.

Here’s how to use it properly.

What AI Is Good At

Finding Information

• Locating relevant sources
• Discovering related topics
• Identifying key authors/papers

Processing Information

• Summarizing long documents
• Extracting key points
• Comparing viewpoints

Organizing Information

• Creating outlines
• Categorizing sources
• Identifying gaps

What AI Is Bad At

Accuracy on Facts

• Hallucinations happen
• Dates often wrong
• Statistics unreliable
• Always verify

Recent Information

• Knowledge cutoffs apply
• Use Perplexity for current data

Original Analysis

• AI synthesizes, doesn’t originate
• Your insights matter most

The Research Workflow

Phase 1: Topic Exploration

Goal: Understand the landscape

Prompt:

I'm researching [topic]. Help me understand:
1. Key concepts I need to know
2. Major debates or perspectives
3. Foundational works to read
4. Current areas of active research
5. Related topics I should explore

What to do with output:

• Use as starting point, not definitive guide
• Note topics to investigate further
• Look up mentioned works to verify they exist

Phase 2: Finding Sources

Tool: Perplexity AI

Better than ChatGPT for finding sources because it:

• Searches the web
• Provides citations
• Links to actual sources

Prompt:

Find academic papers and credible sources about [specific topic].
I need:
- Peer-reviewed research
- Published after [year]
- Focused on [specific aspect]

Tool: Consensus

Specifically designed for scientific research:

• Searches academic papers
• Summarizes findings
• Shows consensus across studies

Tool: Elicit

For literature reviews:

• Finds relevant papers
• Extracts key information
• Helps identify patterns

Phase 3: Reading and Summarizing

For long papers:

Summarize this paper focusing on:
1. Research question
2. Methodology
3. Key findings
4. Limitations
5. How it relates to [your topic]

[Paste abstract and key sections]

For multiple sources:

I have summaries of 5 papers on [topic].
Help me identify:
- Common findings
- Contradictions
- Gaps in the research
- How they relate to each other

[Paste summaries]

Phase 4: Organizing Notes

Creating an outline:

Based on my research on [topic], help me create
an outline for a [paper/report/thesis].

Key points I want to cover:
- [Point 1]
- [Point 2]
- [Point 3]

Sources I'm using:
- [Source 1]: [Main contribution]
- [Source 2]: [Main contribution]

Identifying gaps:

Review my research outline and sources.
What am I missing?
What counterarguments should I address?
What evidence would strengthen my argument?

Phase 5: Writing Support

NOT for writing your paper. But AI can help with:

Clarifying your thinking:

I'm trying to argue that [thesis].
My reasoning is [explanation].
What's weak about this argument?
What would make it stronger?

Improving structure:

This paragraph feels clunky:
[Paste paragraph]

What's wrong with the flow?
How could the structure improve?
(Don't rewrite it, just suggest)

Tool Recommendations

For General Research

Perplexity AI - $20/month (free tier available)

• Web search with citations
• Good for current information
• Better sourcing than ChatGPT

For Academic Papers

Consensus - Free (premium available)

• Searches academic databases
• Shows research consensus
• Great for scientific topics

Elicit - Free

• Literature review assistant
• Paper summaries
• Research workflow tools

For Deep Analysis

ChatGPT Plus - $20/month

• Long context for documents
• Good at synthesis
• Custom instructions for research

Claude - $20/month

• Excellent for long documents
• Strong at nuanced analysis
• Thoughtful responses

For Specific Fields

Semantic Scholar - Free

• AI-powered paper discovery
• Citation analysis
• Influence metrics

Research Rabbit - Free

• Visual paper connections
• Literature mapping
• Collaboration features

Verification Protocol

Every fact needs checking

1. AI states a fact
2. Find the primary source
3. Verify the fact directly
4. Cite the primary source, not AI

Red flags to watch for

• Specific statistics (often hallucinated)
• Paper titles (sometimes invented)
• Author names (can be wrong)
• Dates (frequently inaccurate)
• URLs (often broken/fake)

When AI helps vs. hurts

AI helps: Finding what to look for You do: Actually finding and reading it

AI helps: Summarizing what you’ve read You verify: Summary is accurate

AI helps: Organizing your thoughts You do: The actual thinking

Academic Integrity

Generally Acceptable

• Using AI to find sources
• Summarizing papers for your own notes
• Brainstorming ideas
• Improving clarity of your writing
• Checking grammar

Often Prohibited

• Having AI write your paper
• Using AI summaries as citations
• Not disclosing AI use when required
• Submitting AI output as your own work

Best Practice

• Check your institution’s policy
• Cite AI use when required
• Keep AI to research/editing, not writing
• Make your analysis genuinely yours

Prompts That Work

Understanding a concept

Explain [concept] at three levels:
1. Simple explanation (high school level)
2. Intermediate (undergraduate level)
3. Technical (graduate level)
Include key terminology at each level.

Finding research gaps

What are the current limitations and gaps
in research about [topic]?
What questions remain unanswered?
What methodological challenges exist?

Comparing theories

Compare and contrast [Theory A] and [Theory B]:
- Core assumptions
- Key differences
- Strengths of each
- Criticisms of each
- When to use each

Building arguments

I want to argue [thesis].
What evidence would support this?
What evidence might contradict it?
What are the strongest counterarguments?

Workflow Example

Topic: Impact of remote work on productivity

Step 1: Topic exploration with ChatGPT

• Identified key concepts: productivity metrics, collaboration, work-life balance
• Found major debates: productivity up vs. down
• Noted researchers to look up

Step 2: Source finding with Perplexity

• Found 12 relevant studies
• Verified each exists in Google Scholar
• Downloaded actual papers

Step 3: Paper summaries with Claude

• Uploaded each paper
• Got structured summaries
• Identified patterns across studies

Step 4: Organization with ChatGPT

• Created outline
• Identified where sources fit
• Found gaps needing more research

Step 5: Writing

• Wrote paper myself
• Used AI only to clarify confusing sentences
• All analysis and arguments original

Result: Research that would have taken 40 hours done in 20, with proper sourcing and original analysis.

Common Mistakes

Trusting without verifying

Never cite a paper you haven’t read just because AI mentioned it.

Using AI as a shortcut to thinking

AI helps find information, not understand it. You still need to do the intellectual work.

Over-relying on one tool

Different tools have different strengths. Use several.

Not keeping records

Track what AI helped with. You may need to disclose it.

The Bottom Line

AI accelerates research dramatically when used to:

• Find sources
• Process large volumes of text
• Organize information
• Clarify thinking

But you still need to:

• Verify everything
• Read primary sources
• Do original analysis
• Write with integrity

Use AI as a research assistant, not a replacement for research skills.