Collect the same fields from several pages, combine the rows, then ask Starizon to clean or calculate. Read the result in chat, download a file or send it to a connected spreadsheet.

Choose the pages and fields

Choose AI Chat and select the source tabs or their group with @. You can also ask Starizon to find the pages in your bookmarks or open a list of links before extracting.

For extraction on its own, type /extract and select the source pages. To combine extraction with uploaded files, Knowledge Base material, Spaces or further calculations, describe the whole task in AI Chat.

Name the fields, the source pages and the output:

From the five book-list pages in this tab group, extract book title, author, rating, page count and source URL. Combine the lists, remove duplicate editions, and keep books rated at least 4.2 out of 5 with no more than 300 pages. Give me a CSV and the row counts before and after filtering.

These criteria are examples. Pick conditions that fit your pages, and name the rating scale when sites use different ones.

Clean and calculate

Starizon can run code on the extracted data. You can ask it to:

• Remove duplicates by a key you name, such as ISBN or title plus author.
• Normalize dates, currencies or units, keeping the original values you want to inspect.
• Filter, sort, group and calculate totals or averages.
• Join page data with an uploaded table, keeping unmatched rows visible.

Keep ISBNs and product codes as text so their leading zeros remain. Separate books with missing ratings, and show which source supplied each row.

Give missing values a rule. Say whether Starizon should exclude them, keep them in a separate table or go back to the sources to check again.

Check coverage and results

Check the source pages and row counts in the extraction result. One task can return several tables, such as a product list and a separate specifications table. Review each table before combining them.

Compare the reported sources with the set you asked for. If one page fails, the result can still include rows from the other pages. Ask Starizon to retry the missing pages before treating the table as complete.

When later pages of a list have their own URLs, open them or paste their links, and ask AI Chat to extract from them. Include the page range you want.

If the rows only appear after you click Next, expand a section or use Load more, choose Agent S6 and include those actions in the request. Check which pages it reached before you treat the table as complete.

Spot-check a few rows against their source links, especially identifiers, prices and units.

Deliver or reuse the table

Ask for the complete table in the format you need, such as CSV, and click Download on its file card. The chat may show only some of the rows; check the downloaded file against the reported row count. To write to Google Sheets, connect the app, enable Apps for the conversation and name the spreadsheet or destination.

Send the full filtered table to a new sheet called Reading shortlist in my connected Google Sheets account. Include the source URL column.

Open the link the app returns and check the sheet itself. A table preview in chat does not confirm delivery.

To keep working in another conversation, save the table and conversation to a Space. From there you can ask for new filters or calculations on the saved file.