How I Use AI for Research: Prompts That Actually Work

I used to spend 3-4 hours doing market research before making product decisions.

Now it takes me 30 minutes.

The research is better. More comprehensive. Hits angles I wouldn't have thought of.

The difference isn't the AI. It's knowing how to prompt for research specifically.

Most people ask AI research questions like "Tell me about the market for X." They get a Wikipedia summary that's basically useless.

I'm going to show you the exact prompts I use for:

Market sizing and opportunity analysis
Competitive research that actually helps decisions
Making sense of data and finding patterns
Tracking trends before they're obvious
Customer research without running surveys

These are my actual working prompts, not theoretical examples.

Why AI Research Works (And Where It Doesn't)

First, let's be honest about what AI can and can't do.

What AI is amazing at:

Synthesizing information from many sources
Spotting patterns across datasets
Generating frameworks for thinking about problems
Playing devil's advocate on your assumptions
Organizing messy information into structures

What AI is terrible at:

Knowing things that happened last week
Having proprietary market data
Understanding your specific customers better than you
Making final decisions for you

The key insight: Use AI to structure your thinking and accelerate initial research. Then validate the important stuff with real data.

I've found that AI can handle about 70% of my research work. The other 30%—customer interviews, checking latest data, industry-specific knowledge—still requires humans.

Market Research Prompts

Finding Market Size and Opportunity

This is the first question for any product idea: Is this market big enough to matter?

Here's my actual prompt:

I'm researching the market for [specific product/service].

Break down the market opportunity by thinking through:

(1) Who are the distinct customer segments that would buy this?
(2) Roughly how many potential customers in each segment?
(3) What would they realistically pay?
(4) What's the total addressable market (TAM)?
(5) What percentage is actually reachable (serviceable addressable market)?
(6) What's growing vs shrinking?

Be specific about your assumptions. If you're uncertain, say so and explain your reasoning.

Last month I used this for a SaaS tool idea. The AI broke down the market into segments I hadn't considered (freelancers vs agencies vs enterprises) and sized each one differently.

It also flagged that the enterprise segment was growing 30% year-over-year while freelancer segment was flat. That changed where we decided to focus.

The key detail: "Be specific about your assumptions" forces the AI to show its work. You can challenge the assumptions rather than just accepting numbers.

Understanding Market Segments

Once you know the market exists, you need to understand who's in it.

Segment the [industry/market] into distinct customer groups.

For each major segment, analyze:
(1) What's their core need or pain point?
(2) How do they currently solve this?
(3) What would make them switch solutions?
(4) How price-sensitive are they?
(5) How do they make buying decisions?
(6) What's their typical budget?

Create a comparison table showing how segments differ.

I used this when we were trying to figure out whether to target small businesses or enterprises first.

The AI laid out that small businesses care about price and ease-of-use, decide quickly (owner makes the call), but churn easily.

Enterprises care about security and integration, have long sales cycles (committee decisions), but stick around once they adopt.

Same product, completely different go-to-market strategy needed.

Identifying Market Trends

Markets are always shifting. The trick is catching trends early.

What are the major trends shaping [industry] right now?

For each significant trend:
(1) What's actually changing?
(2) What's causing this change?
(3) Who benefits from this trend?
(4) Who gets hurt by it?
(5) How far along is this trend (early, mid, late)?
(6) What does this mean for a company entering now?

Focus on trends that will matter 1-3 years from now, not just current hype.

The "focus on 1-3 years" is critical. Otherwise you get whatever's trending on Twitter this week.

When I researched AI tools for writing, this prompt surfaced that the trend wasn't just "AI helps writing"—it was "writing tools are becoming more specialized for specific use cases."

That insight led us to focus on a narrow niche instead of trying to be a general writing assistant.

Competitive Analysis Prompts

Direct Competitor Comparison

You need to know not just who your competitors are, but exactly how you stack up.

Compare these companies in the [space]:
- [Competitor A]
- [Competitor B]
- [Competitor C]
- [Our company/concept]

For each, create a profile:
(1) Core value proposition (what they're really selling)
(2) Target customer (be specific)
(3) Top 3 differentiators
(4) Pricing model and typical price point
(5) Main strength
(6) Main weakness
(7) Why customers choose them

Then answer: What's our clearest competitive advantage? Where are we vulnerable?

Format matters here. I ask for a profile of each competitor separately, then a direct comparison.

Last time I did this, the AI pointed out that all three competitors focused on features while customers (based on review analysis) actually cared more about support responsiveness.

That became our differentiation strategy.

Finding White Space

Sometimes the best opportunity is serving a segment everyone else ignores.

Analyze the competitive landscape for [market].

Map the major players by:
- X-axis: Price (low to high)
- Y-axis: Feature complexity (simple to advanced)

For each competitor, explain why they positioned there.

Then identify:
(1) Where is there white space? (underserved positions)
(2) Why isn't anyone there? (is it a real opportunity or a dead zone?)
(3) What would it take to succeed in that position?

This prompt has found me actual opportunities.

When researching project management tools, it showed that everyone clusters at either "super simple" (Trello) or "super complex" (Jira). The middle ground—powerful but approachable—was wide open.

That's where Notion and Linear succeeded.

Win/Loss Analysis

If you're already in market, understanding why you win or lose deals is gold.

Analyze why [our company] wins or loses deals against [competitor].

Based on typical customer decision factors in [industry]:

When we WIN:
(1) What did the customer value most?
(2) What did we offer that competitor didn't?
(3) What was the decision process?
(4) What almost made them choose competitor?

When we LOSE:
(1) What was the deciding factor?
(2) What did competitor offer that we don't?
(3) Was it a feature gap, price, trust, or something else?
(4) What would have changed the outcome?

Then recommend: What's the #1 thing we should improve to win more?

I ask sales teams to give me 5-6 recent wins and losses, feed them into this prompt, and the AI finds patterns I wouldn't have spotted.

Recently it pointed out that we were losing deals not because of our product, but because we didn't have case studies in the customer's specific industry. Easy fix.

Data Analysis Prompts

Finding Patterns in Data

You don't need to be a data scientist to spot meaningful patterns.

Analyze this data for meaningful patterns:
[paste data or describe dataset]

Look for:
(1) Clear trends over time
(2) Unusual spikes or drops (and what might explain them)
(3) Correlations between different metrics
(4) Seasonal patterns or cycles
(5) Outliers worth investigating

For each pattern, explain:
- What the data shows
- What might explain this pattern
- Whether this is significant or just noise
- What I should investigate further

I use this constantly for product analytics.

Last week I fed in user engagement data and asked about patterns. The AI spotted that engagement dropped every Friday afternoon (obvious in hindsight) and spiked on Monday mornings.

That changed when we schedule feature releases and send emails.

Making Sense of Metrics

Sometimes you have numbers but don't know what they mean.

Help me interpret these metrics for [goal]:

[paste your metrics: conversion rates, growth rates, engagement numbers, etc.]

For each metric:
(1) Is this number good, bad, or neutral for [industry/use case]?
(2) What's the typical benchmark?
(3) What might explain why ours is different?
(4) What related metrics should I be tracking?
(5) What action should I take based on this?

Last month our email open rate was 22%. I couldn't tell if that was good or terrible.

The AI explained: For B2B SaaS, 22% is solid (average is 15-20%). But click-through was low (1.2% vs 2.5% average), suggesting the subject lines work but content doesn't engage.

That specific diagnosis led to us rewriting email content, not subject lines.

Cohort Analysis

Understanding how different customer groups behave over time.

Analyze these user cohorts:

Cohort Jan 2024: [retention/usage/revenue metrics]
Cohort Feb 2024: [metrics]
Cohort Mar 2024: [metrics]

Compare across cohorts:
(1) How does retention differ?
(2) What's improving or getting worse over time?
(3) What might explain the differences? (product changes, different customer sources, seasonal factors)
(4) Which cohort is healthiest and why?
(5) What should we do based on this?

When I ran this on our signup cohorts, it showed that users who joined in March had 40% better retention than January.

We dug in and found that we'd improved onboarding in March. That insight made us prioritize onboarding even more.

Trend Analysis Prompts

Evaluating Emerging Trends

Every week there's a new "next big thing." Most are hype. Some matter.

Evaluate this trend: [trend description]

Critical analysis:
(1) Is this a real shift or just hype?
(2) What evidence supports it's real vs overblown?
(3) Who's actually adopting this and why?
(4) What's the timeline? (early adopters now, mainstream in X years, or already peaking?)
(5) What needs to happen for this to become mainstream?
(6) What could kill this trend?
(7) How should [our company/industry] think about this?

Be skeptical. Trends are often overhyped.

The "be skeptical" instruction is key. Otherwise AI tends to be overly optimistic about every trend.

I used this to evaluate whether we should add AI features to our product last year. The analysis showed: real trend, but most users don't care yet. Watch and wait.

We waited 6 months, saw demand actually appear, then built it. Perfect timing.

Understanding Macro Impacts

Big forces (economic shifts, regulation, technology) ripple through everything.

How will [macro trend] impact [our industry/company]?

Analyze:
(1) Direct impact (what changes immediately?)
(2) Indirect impact (second-order effects)
(3) Timeline (when does this hit?)
(4) Who wins and who loses?
(5) What should we start doing now?
(6) What should we stop doing?
(7) What's the biggest risk if we ignore this?

Last year I used this to analyze how rising interest rates would affect SaaS.

The AI connected dots I hadn't: Higher rates → Companies cut costs → Harder to acquire customers → Retention becomes more important → We should invest in customer success.

That analysis drove our whole strategy for the year.

Customer Research Prompts

Building Customer Personas

You need to understand who you're building for.

Create a detailed customer persona for [product/service].

Profile:
(1) Role and typical company size
(2) Daily responsibilities and goals
(3) Main pain points in their work
(4) How they currently solve these problems
(5) What frustrates them about current solutions
(6) What would make them try something new
(7) Budget authority and decision process
(8) Where they go for information and recommendations

Make this specific and realistic, not generic.

The "specific and realistic" instruction matters. Otherwise you get "Marketing Manager at a Medium Company" which is useless.

I ask for personas when launching anything new. Then I validate by actually talking to 5-10 people who match that persona.

Usually the AI gets it 70% right. The interviews fill in the real nuances.

Identifying Pain Points

What problems are painful enough that people will pay to solve them?

Identify the major pain points for [customer type] in [domain].

For each pain point:
(1) What's the problem?
(2) How often does this happen?
(3) What does it cost them? (time, money, stress)
(4) How do they currently handle it?
(5) What would be an ideal solution?
(6) Would they pay to solve this? How much?

Rank pain points by: severity × frequency × willingness to pay

This prompt structure forces prioritization. Not all pain points matter equally.

When researching a product idea, this showed that our assumed main pain point was actually #4 on the list. We pivoted to focus on the real #1 problem.

Saved us months of building the wrong thing.

Advanced Research Workflows

Here's how I actually use these prompts together.

Complete Market Entry Research

When evaluating a new market:

Step 1: Market size and opportunity (first prompt) Step 2: Segment analysis (understand who's there) Step 3: Competitive landscape (map the players) Step 4: Trend analysis (where is this going?) Step 5: Customer pain points (what do they need?)

Output: Decision on whether to enter and what segment to target

This five-step process takes me about 45 minutes now. Used to take 2-3 days.

Competitive Strategy Deep-Dive

When you need to understand your position:

Step 1: Direct competitor comparison Step 2: White space analysis Step 3: Win/loss analysis (if you have deal data) Step 4: Market trend overlay (which trends help or hurt you?)

Output: Your actual competitive advantages and vulnerable points

Product Decision Research

Before building anything significant:

Step 1: Customer persona and pain points Step 2: Competitive alternatives analysis Step 3: Market trends (is this growing or shrinking need?) Step 4: Cohort data if you have it (are current customers actually using related features?)

Output: Build/don't build decision with confidence

The Biggest Mistake People Make

They treat AI research as final answers instead of structured thinking.

AI gives you a framework. A starting point. Hypotheses to test.

Your job is to:

Use AI to structure the research questions
Get initial analysis and patterns
Validate anything important with real data
Talk to actual humans (customers, experts)
Make the decision based on all inputs

The people I see getting the most value use AI to do in 30 minutes what used to take days. Then they spend the saved time validating the insights that actually matter.

How to Validate AI Research

For market sizing: Cross-reference with actual industry reports (Gartner, Forrester, etc.)

For competitive analysis: Read real customer reviews on G2, Capterra Look at competitors' actual pricing pages Talk to people who've used those products

For trends: Check Google Trends for search volume over time Look at which companies are actually raising money in that space Ask people in the industry what they're seeing

For customer research: Interview 5-10 actual potential customers Look at support tickets and sales call notes Run a small survey to test assumptions

AI research is fast. Validation is how you know it's right.

Tools That Complement These Prompts

I use AI for initial research, then validate with:

Google Trends: Confirms trends are real, shows regional differences
Crunchbase: Actual competitor funding and company data
SimilarWeb: Real traffic numbers for competitors
G2/Capterra: Unfiltered customer opinions
Your own analytics: Nothing beats your actual data
Customer interviews: 10 conversations beat 100 AI analyses

Getting Started

If you want to try this approach:

Pick your most important open question right now
Find the relevant prompt above
Adapt it to your specific situation
Run it and see what you get
Validate the 2-3 most important insights
Make your decision

After doing this a few times, you'll develop intuition for which prompts work for which situations.

And your research quality will go up while your research time goes down.

That's the whole point.

These research prompts work even better when you use the right prompting technique for the situation. Check out our guide on types of prompts to understand when to use different approaches.

For complex research decisions, combining this with chain-of-thought prompting makes the analysis even more rigorous.

And if you're comparing AI tools for research specifically, our Claude vs GPT-4 comparison covers which model is better for different types of analysis.

Looking to organize all these research prompts? Read our guide on managing and organizing AI prompts.