Amazon Product Data Extraction with Scrape.do, GPT-4, and Google Sheets - n8n Workflow

• E-commerce store managers needing competitor pricing analysis.


• Market researchers extracting large volumes of structured data.


• Data scientists and analysts looking for reliable web scraping n8n templates.


• Users who require AI-powered data cleaning before saving to a database.


• Developers seeking a robust n8n workflow example for integrating proxy services like Scrape.do.

Extracting clean, structured data from dynamic e-commerce sites like Amazon is challenging due to aggressive anti-bot measures and inconsistent HTML structure. This powerful n8n workflow solves this by combining the robust capabilities of Scrape.do for successful page access with the intelligence of GPT-4o-mini for data transformation.

The process begins by reading a list of target URLs from Google Sheets. For each URL, the n8n node initiates a request via Scrape.do, ensuring a successful page fetch. Crucially, the workflow doesn't just scrape the data; it feeds the raw, messy HTML elements into an OpenAI n8n node configured with a structured output parser. This allows the AI to clean the text, standardize formats (like ratings and prices), and ensure the final output is perfect for analysis. This comprehensive n8n template ensures high data quality and reliability, saving significant manual cleanup time.

This sophisticated n8n workflow operates in seven primary stages:

1. Trigger and Input: The n8n workflow is manually initiated via the n8n trigger node, 'When clicking Test workflow'. It immediately connects to the Google Sheets n8n node to retrieve a column of Amazon product URLs from a specified input sheet.


2. Batch Processing: The 'Loop Through Each URL' n8n node (Split In Batches) handles iteration, ensuring that each URL is processed individually and sequentially.


3. Web Scraping: The 'Scrape Product Page HTML' n8n node uses an HTTP Request to call the Scrape.do API, dynamically injecting the required API token and the current product URL. This bypasses typical scraping limitations.


4. Raw HTML Extraction: The subsequent HTML n8n node extracts specific raw data points (Title, Price, Rating, Review Count) using highly targeted CSS selectors designed for Amazon pages.


5. AI Structuring & Cleaning: The core processing happens in the 'Clean & Structure Data with AI' n8n node (Chain LLM). It receives the raw, extracted elements and feeds them to the configured 'OpenAI Chat Model' (GPT-4o-mini). The prompt explicitly instructs the AI to normalize and structure the data according to the predefined JSON schema (set by the 'Structured Output Parser' n8n node).


6. Output Formatting: The 'Format Final JSON Output' n8n node (Split Out) isolates and flattens the clean, structured JSON object returned by the AI, preparing it for database insertion.


7. Data Storage: Finally, the 'Save Product Data to Google Sheets' n8n node appends the fully cleaned and structured product data (Name, Description, Rating, Reviews, Price) to the designated results sheet in Google Sheets.

To deploy and utilize this powerful n8n workflow template, follow these setup steps:

1. Import the n8n Workflow: Copy the provided JSON code and import it into your n8n instance.


2. Set up Credentials:

1. Configure Workflow Variables (Essential):

1. Prepare Google Sheets: Create a spreadsheet with at least two tabs: one for input URLs (with a column header 'url') and one for results (the target sheet for the 'append' operation).


2. Test the n8n trigger: Execute the 'When clicking Test workflow' n8n trigger to ensure all connections and scraping steps are successful.

Manual Trigger (n8n trigger): Starts the n8n workflow manually for testing or immediate execution.
Google Sheets (1. Get Product URLs): Retrieves data from the specified input sheet ('Sheet1') using the document ID 19Allmozbygw-QogPeq2TH9m9D57FCn4MTu3zmJukg1A. This n8n node supplies the list of URLs.
Split In Batches (2. Loop Through Each URL): A core n8n node for iteration, it ensures that subsequent operations process each URL sequentially rather than trying to handle all requests simultaneously.
HTTP Request (3. Scrape Product Page HTML): Makes the request to the external Scrape.do API. Configuration includes injecting the workflow variable SCRAPEDOTOKEN and the current URL item, making this the critical scraping step.
HTML (4. Extract Raw Data Elements): This n8n node uses specific CSS selectors (#productTitle, .a-price .a-offscreen, etc.) to pull raw, unformatted text chunks from the scraped HTML content.
Chain LLM (5. Clean & Structure Data with AI): The primary AI cleaning n8n node. It uses a detailed prompt and the connected GPT-4o-mini model to convert the raw, disparate HTML elements into a coherent JSON structure.
OpenAI Chat Model (GPT-4o-mini): Configured with a low temperature (0) for reliable, deterministic output and the jsonobject response format, maximizing data consistency for the n8n workflow.
Split Out (6. Format Final JSON Output): This utility n8n node takes the nested 'output' JSON object generated by the AI and promotes its contents (name, description, rating, reviews, price) to the main item level, making them directly accessible for the final storage step.

• Google Sheets (7. Save Product Data): Performs an 'append' operation to the output sheet ('Sheet2'), storing the final, clean product data extracted by the n8n workflow.