ScrapeSheet has two ways of getting data out of a page. The regular extractor spots tables and repeated lists instantly and for free. The AI extraction is for the pages where that isn't enough: details scattered through text, cards with an odd layout, or when you want specific fields with clean names. It uses your own API key (BYOK — bring your own key), so there is no ScrapeSheet fee per row or per page.
Step by step
- Create an API key with Anthropic (Claude) or OpenAI. They bill you directly for what you use.
- Open the page, open the ScrapeSheet side panel and expand 🤖 AI for tricky sites. Paste your key: it is stored only in your browser.
- Type — or dictate with the microphone — what you want. Be specific about the fields:
every event with date, venue, city and ticket link,each plan with its name, monthly price and storage. - Press Extract with AI. Review the table (you can delete columns with ✕) and export to Excel, CSV or Google Sheets.
When to use AI, and when not to
| Regular extractor | AI extraction | |
|---|---|---|
| Best for | Tables, product grids, result lists, directories | Text-heavy or irregular pages, specific fields, cleaning up |
| Speed | Instant | A few seconds per page |
| Cost | Included (free plan limits apply) | Pro + your AI provider's usage fees |
| Many pages | Yes, up to 200 per run (Pro) | The page you are on |
The AI reads the visible text of the page; very long pages are trimmed. For a list with hundreds of items, use the regular extractor with pagination and keep AI for the awkward pages. As with any AI, check the result before relying on it.
Privacy and cost
- Your key stays in your browser. The page text goes from your browser directly to your AI provider; it does not pass through ScrapeSheet's servers.
- By default ScrapeSheet uses low-cost models (Claude Haiku or GPT-4o mini); you can set another model in Settings.
- AI extraction is part of ScrapeSheet Pro. The provider's fees are separate and billed by them.
Use it responsibly
ScrapeSheet is meant for publicly available data. It reads the page you have open in your own browser, at your own pace; it does not get around logins, paywalls, CAPTCHAs or anti-bot protection, and it should not be used to try. Before you extract, check the website's terms of use: if a site forbids automated collection, don't use ScrapeSheet there. If the data includes people (names, emails, phone numbers), privacy law such as the GDPR applies to what you do with it. More in Is web scraping legal?
FAQ
What is a BYOK AI web scraper?
BYOK means bring your own key: you paste your own Anthropic or OpenAI API key into ScrapeSheet, the AI runs on your account and your provider bills you directly. ScrapeSheet adds no per-row or per-page AI fee.
Which AI models does ScrapeSheet work with?
Claude models from Anthropic and OpenAI models such as GPT-4o mini. By default it picks a low-cost model; you can type another model name in Settings.
Does my data go through ScrapeSheet's servers?
No. The page text is sent from your browser directly to the AI provider you chose, and your key is stored only in your browser.
Do I need the AI to scrape a normal table or list?
No. The regular extractor detects tables and lists on its own without AI. AI is for irregular or text-heavy pages.
Is AI extraction free?
AI extraction is a Pro feature, and your AI provider charges you for usage on your own account.
Related use cases
Back to the ScrapeSheet home page · all use cases · ScrapeSheet Pro