How BuiltWhen estimates website history
BuiltWhen does not claim to know the exact construction date of every website. It estimates website history using observable evidence, and shows that evidence so you can judge it yourself.
What BuiltWhen measures
Domain registration
The date the domain’s current registration began, from public registry data (RDAP). A website can’t launch before its domain exists, so this is a lower limit. It is never reported as the website’s launch or age.
First known evidence
The earliest copy of the homepage in the Internet Archive that shows a real website. Parking pages, “coming soon” pages and default server pages are recognised and skipped — the last of those becomes a lower limit instead.
Previous websites on the same domain
Domains often change hands. When the archive shows an unrelated website earlier on — different layout, different title, and no shared wording with today’s site — BuiltWhen treats it as a previous owner’s site: its history isn’t counted, the timeline marks the switch, and confidence is lowered slightly because this is a judgement from page content, not a record of ownership.
Estimated launch
The month of the earliest reliable evidence, with a window reaching back to the latest lower limit (at most six months). Weaker, earlier signals such as a copyright year or a sitemap date can move the estimate only when an independent second source agrees. A copyright notice showing only the current year is ignored — it’s usually generated automatically.
Technology history and website changes
Up to six archived copies of the homepage, spread across the site’s history, are compared. A large change in page structure is reported as a possible redesign; a change of platform (for example WordPress to Webflow) as a migration. Dates are narrowed with a few extra archived copies and always shown as a range. If the Internet Archive can’t be reached, the report says so rather than guessing.
How confidence is scored
Each source type has a weight. Evidence earns points when it agrees with the estimate; only the best item per source type counts, so one source can’t vote twice. Points are compared against a reference total, then reduced when signals conflict, when the domain looks re-registered, or when only a year is known.
- domain registration: 10
- archive: 30
- search index: 20
- technology: 15
- content date: 10
- external reference: 10
- http metadata: 5
- sitemap: 10
- website metadata: 10
High confidence means several independent signals agree. Medium means there’s good evidence but little corroboration. Low means the evidence is thin. When nothing reliable is found, BuiltWhen says “unknown” rather than falling back to domain age.
What BuiltWhen does not guarantee
- The exact date development started or the site went live
- The exact date of a redesign
- Who owns or built the website — registrant data is never collected
- Private infrastructure, staging sites, or anything not publicly visible
- Changes that happened between archive captures, or on sites the archive never captured
How BuiltWhen fetches websites
It reads robots.txt and respects it, fetches only the homepage and one sitemap, identifies itself as BuiltWhenBot, uses strict timeouts and size limits, and never connects to private or internal network addresses. Archive lookups are cached so popular sites aren’t fetched repeatedly.