Contact discovery

When the side panel opens on a company's site, the content script extracts every contact-page link, phone number, and email address it can find from the current page (and the linked contact pages). They surface in the panel as a "How to contact" card with one-click actions.

What gets extracted

• Phone numbers — recognized with a permissive regex that handles US, international, and toll-free formats. Each phone number gets a one-click tel: action.
• Email addresses — every mailto: link plus heuristically-detected email patterns in plain text. One-click mailto: action.
• Chat widgets — if the page has an embedded chatbot (Intercom, Drift, etc.), a one-click Open chat button surfaces it without you having to hunt for the widget bubble.
• Contact-page links — anything the page advertises as a contact route (/contact, /support, /help, etc.) shows up as a clickable link.

Confidence and ranking

The extracted contacts are ranked by:

1. Source proximity — numbers found on a /contact page outrank numbers in the page footer.
2. Format quality — 1-800-… formatted numbers outrank bare digit strings.
3. Verified Call Tree match — if a number matches a known node in the company's Verified Call Tree, it gets a green check and surfaces first with the department label.

What never shows up

• Numbers and emails inside form input fields (which would expose data the user is typing).
• Anything from pages behind a sign-in (the content script doesn't read authenticated pages unless the panel is open and you've explicitly followed the company).
• Anything from <script> blocks or stringified JSON in the page source.

Related

• Verified Call Trees → — the trusted alternative to scraped numbers.
• Auto-detection → — what else the content script does on first visit.