Nextbrowser
Download
Use cases0:50 watch

Extract Amazon reviews with NextBrowser

Collect accessible Amazon reviews in a NextBrowser session, check product matching and export a deduplicated CSV with honest collection limits.

About this reference video

The video shows the source task and its recorded outcome. The guide below explains how to run that task through NextBrowser: choose a profile and browser toolset, start a session, then supervise your local agent in Chat and Live View. Interface labels and results can differ from the recording.

Go to the NextBrowser workflow ↓
Published · @AIWebAutomationLabWatch on YouTube ↗

Overview

Amazon review access and product matching can limit an extraction before analysis begins. In NextBrowser, confirm the intended storefront and products, then ask the local agent for the reviews actually available in that session. Keep access failures visible so the exported dataset does not imply coverage it never achieved.

What happens in the video

The agent checks three Amazon products, corrects product-identifier mismatches, and encounters sign-in requirements on review pages. Its final message reports 37 review rows collected from visible product-page reviews, and a CSV is opened on screen. The published description mentions approximately 300 rows, but the recording reports 37.

Before you start

  • NextBrowser on macOS or Windows, signed in through browser pairing
  • An installed and authenticated local agent: Claude Code or Codex from the ChatGPT desktop app
  • Exact product URLs and a review schema with rating, title, date, text and product URL

Run this in NextBrowser

Follow these steps in your own NextBrowser workspace. The video summary records the source demonstration; your result needs the checks below.

  1. Prepare the NextBrowser session

    Select a profile for the intended Amazon region, start it and check the storefront and products in Live View. If full reviews require sign-in, decide whether to sign in yourself or limit the run to public content.

  2. Give the local agent a scoped task

    Select your local agent in NextBrowser. Confirm the active profile and tab for the supplied Amazon product URLs. Submit the prompt below in Chat after replacing its placeholders. The selected profile name is guidance for the agent; check the actual session in the sidebar and Live View.

  3. Collect the task evidence

    Ask the agent in Chat to verify each product identifier, open accessible review sections and collect one review per row. Keep product URLs, missing values and pagination failures in the output instead of substituting another product.

  4. Review the result against the browser

    Open the CSV produced by the agent and reconcile its row count with the run log. Check duplicates, multiline text and a sample against the browser. The recording reports 37 reviews; that is not a target or a guaranteed yield for your session.

Send this prompt in NextBrowser Chat

Suggested prompt

Use the intended NextBrowser profile [profile name]. First confirm the running session, active tab and target site; stop if they do not match. Collect public reviews for [three Amazon URLs]. Export product URL, rating, review title, date, text, and available metadata to CSV. Deduplicate, keep one review per row, and report missing fields and the collection count. Report source URLs, the actual amount collected and any incomplete checks. Keep this run to research; do not purchase, publish or change account settings.

Replace the placeholders, then send this in NextBrowser Chat with the intended profile and agent selected. This is a suggested task, not the recording transcript.

What to verify in your run

  • A CSV of the reviews actually accessible in your session, with product provenance
  • An extraction log with actual row counts, missing fields and any sign-in or pagination limits

Questions about this workflow

Does the recording collect 300 reviews?

The YouTube description says approximately 300, but the final message in the recording reports 37 rows. Full-review pagination requires sign-in in that session, so the guide uses the recorded result.

Is extraction sentiment analysis?

No. This recording focuses on structured data collection. Sentiment and theme analysis are subsequent tasks.

Keep exploring

NextBrowser documentation

NextBrowser documentation ↗Product guide: profiles, agents and sessions ↗

From the blog

How to Scrape Reviews from Amazon: Full Guide 2025 ↗

Explore the review extraction workflow and its constraints.

Related use cases

Web Scraping and Browser Research ↗

NextBrowser workflow adapted . Recording reviewed . Video source: published video and description.

More workflows to explore

← Browse all demos