AI can collect product reviews into a sourced report your team can inspect.

Define the review universe before you start collecting.

“Reviews” can mean App Store ratings, G2 comments, Reddit threads, retailer feedback, or posts on a marketplace. They differ in audience, date range, moderation, and whether a rating is attached.

Choose the sources, date window, product variants, and maximum number of reviews before the collection starts. That makes the resulting report explainable and lets the next run compare like with like.

Preserve the evidence before you summarize it.

A chart that says “customers dislike setup” is not enough.

The report should let a product manager open the reviews behind that conclusion and see whether they describe onboarding, hardware installation, billing, or something else.

Strawberry can work through the public pages you select and prepare a structured dataset before producing the narrative. Keep original wording available alongside any short summary.

Separate recurring signals from loud one-off complaints.

One vivid review may be important, but it is not automatically a trend.

Group similar comments, count the collected mentions, and report the uncertainty where language is ambiguous or the sample is thin.

A useful weekly or monthly report can distinguish repeated friction from edge cases, while still reserving a section for severe individual issues that need attention regardless of frequency.

Turn the report into a product conversation, not a scorecard.

The next action may be a support macro, a documentation fix, an experiment, or deeper user research. Attach a proposed owner and a question to each major theme so the team can decide what to investigate next.

Save the collection method as a skill if the sources and taxonomy are stable. A routine can prepare the fresh evidence and comparison each week without quietly changing the source set.

Frequently asked questions