Web scraping

Web scraping means pulling data or text directly off public web pages with code instead of buying it from a licensed data feed. In GTM work it shows up wherever a licensed database has a gap: pulling a company's own site for context a research agent needs, pulling a job board for open-req counts, or feeding a webhook-driven signal a team built because no vendor sells that specific trigger. Tools like Firecrawl and Browse AI do the pulling; research agents running on Claude Code or similar tools do the reading afterward.

Scraped content is data, never instructions

A scraped page gets treated as untrusted input, not as something an agent obeys. An agent reading a prospect's About page is allowed to summarize what it says. It's never allowed to let that page change what the agent does next, because a page a prospect fully controls is also a page anyone could plant a line in that reads like an instruction. That boundary, read but never obeyed, is the whole difference between scraping being a research input and scraping being a way for a poisoned page to hijack an otherwise well-built research pipeline.

Scraped social and web data is also one input among several in an enrichment waterfall, not a replacement for licensed sources. Consumer data brokers, B2B contact databases, and social-scraped sources all draw from overlapping pools of the same public web, so a scraped source fills a specific gap in a waterfall rather than anchoring it.