Web Intelligence

Web Intelligence lets your SmartForms and CoPilots pull fresh context from the web. Use it when your tool needs current facts, or when users paste a URL.

Pick the right feature

Web Searching (Internet Access)

Use this when you need quick, up-to-date facts. It pulls information from the main Google results page. It does not click into results or scrape pages.

Good for:

  • “What’s the latest news about ___?”

  • Company snapshots, basic public info, quick research

  • Reducing outdated answers and hallucinations

Web Crawling (URL scraping)

Use this when you need the actual text from a specific page. Your tool crawls a single URL and brings that page’s text into the prompt.

Good for:

  • Summarizing an article someone links

  • Turning a blog post into email/social variations

  • Extracting details from a product or landing page

Web crawling is available on GPT‑4-class models and above.

Best practices

  • Ask for sources. Have the tool cite the URL or query it used.

  • Keep tasks narrow. “Summarize this URL in 5 bullets” beats “analyze everything.”

  • Verify before acting. Treat web results as input, not truth.

Known limits (so you can design around them)

  • Web Searching only sees the search results page.

  • Web Crawling reads text content from a URL. It may skip paywalled, blocked, or heavy JS-rendered pages.

  • Both features can fail if the source blocks bots or rate limits.

Common patterns

  • SmartForm “Research + Write”

    • Input: topic or brand name

    • Web Searching: pull current context

    • Output: brief + draft content

  • CoPilot “Paste a link, get help”

    • User drops a URL

    • Web Crawling: ingest the page

    • CoPilot: answers questions using that page as context

Safety and compliance notes

  • Only crawl content you’re allowed to access and reuse.

  • Avoid collecting sensitive personal data from the web.

  • If you sell outputs, add your own review step.

Last updated

Was this helpful?