Web page listings pulled into Google Sheets with Jina and AI

Web page listings pulled into Google Sheets with Jina and AI

An n8n workflow that reads one web page through Jina's Reader, asks an OpenAI chat model to pick out every listing on it
and adds one row for each listing to a Google Sheet. The file comes set to a category page of a practice bookshop.

What is inside

• workflow.json: the workflow as n8n exports it, seven nodes: six that do the work and one sticky note. It holds no
  credential.
• README.md: these notes.

What we changed from the original copy

• The credential links are gone: the Google Sheets account, the OpenAI key and the Header Auth credential on
  Jina Fetch. Select your own.
• Save to Google Sheets pointed at the author's own sheet and its Sheet2 tab. It now has YOUR_SHEET_ID and the tab
  Sheet1.
• The schema in Information Extractor held only the results list, with no outer type and properties, so it was
  not a complete JSON Schema. Mindket's copy wraps the same fields in them.
• The column list in Save to Google Sheets named name, price, availability, image and link, while the
  extractor returns title, price, availability, product_url and image_url. It now names the extractor's own
  fields, and it no longer lists a matching column.
• Jina Fetch had its option to ignore SSL issues turned on. It is off now, so the request checks the server's
  certificate.
• The large note at the top, with the author's tutorial video and a link to a sample sheet, is now a title.
• Nothing else.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• A Google Sheet with these headers in row 1 of its first tab: title, price, availability, product_url and
  image_url. The tab is called Sheet1 unless you renamed it.
• An OpenAI API key.
• Optional: a Jina key. See Good to know.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add a Google Sheets OAuth2 credential and select it in Save to Google Sheets. Replace YOUR_SHEET_ID with your
   sheet's ID, the long string in its address between /d/ and /edit. Change the tab name if yours is not Sheet1.
3. Add an OpenAI credential and select it in OpenAI Chat Model. The node names no model, so choose one.
4. In Jina Fetch, either add a Header Auth credential with your Jina key, or set Authentication to None to run without
   one. The node asks for a credential until you do one or the other.
5. Check that row 1 of your sheet holds the five headers above.
6. Click Test workflow. A row for each listing on the page appears in your sheet.

To read another page, replace the address in the URL of Jina Fetch. Keep https://r.jina.ai/ in front of it.

How a run goes

1. When clicking "Test workflow" starts it by hand.
2. Jina Fetch asks Jina's Reader for the page. It comes back as plain text, which n8n keeps in data.
3. Information Extractor sends that text to the chat model with a system prompt and the schema. It returns output,
   an object with a results list. Each book in it has a title, price, availability, product link and image link.
4. Split Out turns output.results into one item for each book.
5. Save to Google Sheets appends each item as a row. Fields go to the headers with the same name.

Change the look

• The page: the URL in Jina Fetch.
• What is picked out: the system prompt and the schema in Information Extractor, and the headers in your sheet to
  match.
• The model: OpenAI Chat Model.
• The sheet and its tab: Save to Google Sheets.
• How it starts: replace the manual trigger with a Schedule Trigger to run it on a timer. Each run adds rows.

Good to know

• Every run appends. A second run on the same page adds the same rows again, so clear the old ones first if you do not
  want duplicates.
• It reads one page. There is no paging, and a page whose text is longer than the model accepts will fail.
• A field that has no matching header becomes a new column on the right, by default, in n8n's current source read on
  12 November 2026. Keep the headers spelled as above.
• The model reads the text and can slip, and every value is text. Check some rows against the page before you rely on
  them.
• The OpenAI Chat Model node, at the version this file uses, names no model and defaults to gpt-5-mini in n8n's current
  source, read on 12 November 2026. OpenAI's deprecations page gives 11 December 2026 for gpt-5-mini and names
  gpt-5.6-terra as its replacement.
• The page the file reads, books.toscrape.com, calls itself a fictional bookstore made to be scraped, and its prices
  have no real meaning. Check a site's rules before you point the workflow at another one.
• Jina's Reader page, read on 12 November 2026, says a key is optional and that without one the limit is 20 requests a
  minute for each IP address. That page does not show the header's name, so take it from Jina's documentation.
• The page's text goes to OpenAI, and Jina sees the address you ask for.