A guide to the public sources that actually matter for company research — filings, registrations, and records — and how to read them.
Most useful company research doesn't come from proprietary databases — it comes from public records that are freely available but scattered across a dozen different sites. Knowing which ones matter, and what they actually tell you, is most of the work.
Every company is registered somewhere — a state Secretary of State filing in the US, Companies House in the UK, equivalent registries elsewhere. These filings tell you the legal entity name (often different from the brand name), incorporation date, registered agent, and officer names. This is the most reliable "does this company actually exist as described" check, and it's usually free to search.
Depending on the sector, there's often a regulator with public filings: SEC filings for anything approaching a public offering, FCC filings for telecom, FDA submissions for health and medical products, patent filings for anything IP-heavy. These are dense but specific — a single patent filing can confirm or contradict a company's claimed technical differentiation faster than any pitch deck.
Standard practice is searching for what's been written about a company. Less standard, but just as useful: noticing what hasn't been written. A company claiming years of traction with zero press mentions, no product reviews, and no conference presence is a signal worth weighing, even though "no news" isn't itself disqualifying.
Beyond the marketing copy, a company's own site tells you things indirectly: job postings reveal what they're actually building and hiring for next, a leadership or "about" page (or its absence) tells you how much they want scrutiny, and a careers page with stale postings is a mild signal worth noting.
WHOIS-style domain registration data and basic infrastructure lookups can confirm how long a company has actually operated online versus how long they claim to have existed — useful for catching a rebrand or a shell that's newer than its "founded in" claim suggests.
None of these sources is disqualifying on its own — a young domain doesn't mean a company is fake, and a missing news mention doesn't mean it's failing. The value is in cross-referencing: does the registration date match the "founded" claim, does the funding history match the headcount implied by job postings, does the competitive positioning hold up against what's publicly documented about the named competitors.
Doing that cross-referencing by hand, source by source, is exactly the manual work that a synthesis-first research tool is built to remove — pulling from these public sources automatically and surfacing the parts that are actually worth your judgment.