How to Prompt AI as a Researcher

TL;DR: Research takes too long when you're doing all the reading yourself. AI can compress most of it by going and reading the sources you would have read, then pulling out what matters. The key is being specific about what you need at the end, not just the topic.

Research takes too long. McKinsey found that knowledge workers spend nearly 20% of their workweek just searching for information. Building a market overview, a competitive landscape, or a due diligence brief the proper way means hours of reading, cross-referencing, and synthesising sources that don't agree with each other.

AI can compress most of that. Not by guessing, but by actually going and reading the sources you would have read, then pulling out what matters.

What can AI actually do for researchers?

Real research tasks: building market overviews, mapping competitors, pulling insights from documents and reports, analysing customer sentiment from reviews, identifying experts and investors in a space. The AI doesn't replace your judgement on what findings mean, but it can do the reading.

What we've found changes output quality most is sending the AI to explore before it asks questions. Before building a competitor map, it checks who the main players actually are. Before summarising a market, it scans what's been published recently. With Strawberry, your companion runs those sweeps first and comes to you with a more informed set of questions, rather than a generic summary based on what it already knew.

What's the difference between a useful AI research prompt and a bad one?

A bad research prompt is too abstract: "Tell me about the project management software market." The AI produces a generic overview that's accurate but useless.

A good prompt defines the scope, specifies what you need to walk away with, and tells the AI how to structure the output so it's actually usable. The best prompts also tell the AI to go find things, not just recall them. Strawberry automatically stores context about your ongoing research projects in memory, so your companion builds on what it already knows each time rather than starting fresh.

What are the best AI prompts for research?

These are the prompts we see working consistently for research workflows using Strawberry.

Building a market authority report

Help me build a market authority report. Ask me which company or market to focus on. You might want to form your own view on what signals authority in that space before asking me. Then run the research and generate a structured report.

That second sentence changes the output noticeably. You're asking the AI to think before it asks, which means the question it eventually asks is more useful.

Pulling insights from reports across your tools

Help me pull insights from research reports across my tools. Ask me which tools to check and what topic or question I'm trying to answer. Go through the relevant documents and synthesise the key findings into a structured summary.

This is the prompt for anyone drowning in PDFs. The AI reads them, you read the summary.

Mapping a competitor landscape

Help me map a competitor landscape. Ask me which company or product to benchmark against and what dimensions matter most. You might want to explore the space yourself first to see who the main players are before asking me. Then structure the output as a comparison I can share with a team.

The "explore first" instruction produces noticeably better output. The AI comes back with a more informed set of questions.

Tracking news and signals across a topic

Help me track news and signals across a specific topic or market. Ask me what to monitor and how often to check. Then set up a Routine that collects the relevant updates and sends me a digest on a schedule.

This turns a one-off research task into an ongoing feed.

Running background research on a person

I need background research on someone. Ask me who and what the context is. Then find their LinkedIn, recent public activity, any press mentions, and relevant professional history. Give me a structured brief.

How do you get better research output from AI?

Be specific about what you need at the end. "Research X" is not a task. "Research X and give me a three-paragraph summary, the five most important players, and any gaps in the data" is.

Let the AI decide where to look. Telling it to "go find what's relevant" usually produces more thorough output than spelling out exactly where to search. It tends to find sources you wouldn't have thought to check.

If you're doing regular research on the same topic, build context over time. Tell your companion what you already know, what you're trying to understand, and what good output looks like. It gets faster and more accurate the more it knows.

We've written similar guides for scraping and data collection and operations teams if those workflows are relevant to you.

What's the best AI tool for research?

Research requires an AI that can actually browse the web, read documents from your Drive, and structure findings. Not Wikipedia summaries.

Strawberry Browser is an AI browser where your companion can run research tasks in the background, pull from your existing documents, and deliver structured output directly into your workflow.

Try it at strawberrybrowser.com.