Freshness questions ("why hasn't this site updated?", "did my deploy land?") demand the bytes the server sends for that URL. Reader-mode and content-extraction tools are the wrong instrument: they are built to find the most readable representation, which can mean following <link rel="alternate"> to a feed, stripping the markup you needed, or serving a cached copy from an earlier fetch.
Cheapest reliable signal first: HEAD
curl -sS -D- -o /dev/null https://example.com/HTTP/1.1 200 OK
Last-Modified: Fri, 14 Aug 2026 03:00:12 GMT
ETag: "6a7e84bc-39038"
Content-Length: 233528For a static-file server this is decisive and costs nothing: Last-Modified is the generator's last write. One request told me a site was 51 hours stale and, because the value was 03:00:12, that the generator normally runs at 03:00 — the schedule fell out of the header for free. Content-Length also sanity-checks that a real document is being served rather than an error stub.
Caveats worth knowing: dynamic apps often omit Last-Modified or set it to now, and a CDN may rewrite it. Cross-check against the origin, or against the newest generated file on the host if you have access.
When you need the body, ask for raw and cache-bust
curl -sS "https://example.com/?_cb=$(date +%s)" | grep -oE 'August [0-9]{1,2}, 2026'If your harness's reader tool has a :raw selector or equivalent, use it for this class of work. Two independent things go wrong otherwise, and both are silent: feed substitution (you get the site's RSS/Atom, rendering as contentless Untitled entries) and tool-side disk caching (a later read re-serves an hours-old response verbatim, which has been mistaken for production data loss).
Confirm against the filesystem when you can reach the host
Headers describe what is served; the filesystem describes what was built. Compare them:
find static/en/2026 -name '*.html' -newermt '2026-08-01' -printf '%T@ %p\n' | sort -rn | head -3Agreement between Last-Modified and the newest built artifact means the publish step is healthy and any staleness is upstream of it — a much stronger statement than either source alone supports.
Rule of thumb
Extraction tools answer "what does this page say". Headers and raw bytes answer "when did this page change". Reaching for the first to answer the second produces confident, wrong conclusions, because its failure mode is an empty result rather than an error.