Automate web research reports by making the sources do the work, not the summary.

Start with the decision, not a pile of links.

“Research this market” can mean a pricing scan, a competitor comparison, a customer list, a policy update, or a technical due-diligence question. Define who will use the report, what they need to decide, which geography or time period matters, and what evidence would change the answer.

That scope tells Strawberry what to look for and what to leave out. A narrow question produces a report someone can use.

Gather sources in layers and check the originals.

Search can find the right trail, but a snippet may be stale, incomplete, or attached to the wrong page. For material facts, open the original company page, filing, documentation, data release, or primary publication and capture the relevant passage.

Use secondary reporting to add context and identify contradictions, not to replace the original evidence where it is available.

Write findings that retain their confidence level.

A good report does not make every sentence sound equally certain.

Label what the source states, what the evidence suggests, and what remains unknown or time-sensitive.

Ask Strawberry for a table of claim, evidence, source date, and confidence before the narrative summary. This makes it easier for a reader to challenge or update one conclusion without redoing the whole research pass.

Make recurring intelligence easier to trust.

A weekly competitor or sector brief only works if it checks the same relevant sources, records what changed, and does not recycle an old conclusion as new. Save the source list, inclusion rules, and output format as a research skill.

A routine can refresh that defined watch and prepare a change report. It should surface uncertainty and source failures rather than quietly filling gaps with plausible text.

Frequently asked questions