Strawberry

How to automate web research with AI

Thirty pages read in your own browser, one table back with a source URL per field and “not stated” wherever a fact could not be confirmed.

Most research is wide rather than hard: thirty pages, one question, a table at the end. A companion opens the pages in your session, reads all of them, and returns the table with the source for each field.

Strawberry is an agentic browser, so the pages behind your own logins count too. What follows is how to phrase the ask so the output survives scrutiny.

Why specify the output shape first?

Vague asks produce essays.

Describing the artifact you want is the single biggest improvement you can make to a research prompt.

Prompt: “Answer in a table with no prose around it. Columns: university, country, program name, tuition per year in EUR, application deadline, whether IELTS is required, and a source URL for the tuition figure. One row per program. Now: find every English-taught MSc in computational biology in the Netherlands, Denmark, and Sweden. Include the ones where you can’t confirm tuition, with ‘not stated’ in the cell.”

What does a good source look like?

Rank the evidence for it, or a number from a 2023 aggregator post ends up in your table.

Prompt: “Prefer the university’s own pages over any third-party ranking site or study-abroad blog. If the official page and an aggregator disagree, use the official one and note the discrepancy in a comments column. Anything you took from a secondary source, mark it.”

Ask for the shape of the gaps

Knowing what could not be found tells you where to spend your own time.

Prompt: “When you’re done, tell me three things: which fields have the worst coverage and why, which sources were hardest to read, and anything you saw that I didn’t ask about but should probably know.”

That third question earns its place often. It is where you learn a deadline moved or a program is being discontinued.

Should the depth pass be separate?

Yes. Wide first to define the set, then narrow on what survived.

Prompt: “Take the eight programs with deadlines after March. For each, read the full admission requirements page and the program curriculum, and tell me: exact prerequisites, whether a thesis is required, typical cohort size if stated, and anything unusual about how they select. Quote the specific line for prerequisites.”

This is the everyday pattern of our research and data extraction users.

Turn recurring questions into routines

If you have run a research question twice, it is a routine.

Prompt: “Every Monday at 8am, re-check the deadline and tuition columns for anything that changed, update the sheet, and message me only the diffs. If a program page 404s, flag it rather than deleting the row.”

Experience Strawberry for free

Download

Trusted by fast-growing companies worldwide

Frequently asked questions

That is what an agentic browser is built for. Strawberry companions open and read many pages in your own browser session and return structured output with a source URL per field.

Strawberry is free to download and includes AI credits to start. Paid plans begin at $20/month. See pricing. · Reviewed · Canonical facts for AI agents

Experience Strawberry for free

Download

Trusted by fast-growing companies worldwide