HTTP 200 says the request succeeded at the protocol level, not that the page users need is working. Add a content assertion: a stable required marker must appear in the final GET body, while a known failure string should not. First confirm the marker is in server-delivered HTML rather than inserted only by JavaScript.

A hosting provider can serve a maintenance placeholder with 200. A CMS can return an empty body or exception text under the same status. A status-only monitor treats all of them as success. RFC 9110 defines 200 as a successful HTTP request; it cannot judge whether a checkout button works. Monitor content on a user-critical path. In enterno this is configured on an HTTP monitor.
Four HTTP 200 responses, one useful page
We ran a local HTTP fixture with four responses. All returned 200; only the healthy case contained sku-ready. These are controlled examples, not customer incidents or a production prevalence sample. The fixture is in method.py; captured results are in data.json.
| Case | HTTP | Body bytes | sku-ready | Content verdict |
|---|---|---|---|---|
| Healthy product card | 200 | 82 | present | healthy |
| Blank response | 200 | 0 | absent | failed |
| Hosting placeholder | 200 | 72 | absent | failed |
| CMS error page | 200 | 59 | absent | failed |
Local fixture summary from data.json
/healthy HTTP 200, body 82 bytes, sku-ready: yes
/empty HTTP 200, body 0 bytes, sku-ready: no
/placeholder HTTP 200, body 72 bytes, sku-ready: no
/cms-error HTTP 200, body 59 bytes, sku-ready: no
For your site, inspect the body with curl -sS -L https://example.com/. Use -L only if the monitor also follows redirects; compare the final page in either case. If a slow backend causes a placeholder, inspect the first-byte and total timings as well as body content.

Choose a marker that proves the page works
A good marker appears only on the useful page and survives ordinary copy changes. A product card may have a stable server-rendered attribute; a dashboard might expose a small status label in its HTML. Avoid a word that appears in the shared footer or a hidden script comment. enterno searches the raw response body case-insensitively using mb_stripos(). It does not run JavaScript or judge whether the marker is visible to a user. A client-side-only element therefore needs a browser-based or dedicated API check instead.
The not_contains mode can detect a stable error such as Fatal error, but absence of that string does not prove the page is useful. For a critical path, pair a required healthy marker with a forbidden error marker using separate monitors when appropriate. A keyword belongs on an HTTP monitor; enterno API monitors use assert_body and reject the HTTP keyword field. Before testing the body, make sure the probe is allowed through the WAF. The launch checklist maps these checks to risks.
Inspect redirects, cache and authentication
Test without your browser session. An authenticated dashboard may redirect a monitor to the login form. Do not paste a private cookie into a public article; use protected request settings and a limited test account where appropriate. If a CDN caches a placeholder, compare the public route with the origin. If a proxy returns 502 instead, locate that hop using the 5xx guide.
Watch for the opposite failure: an editor changes the marker while the feature still works. A stable machine endpoint or API assertion is better than marketing headline copy. Google SRE emphasizes user-visible symptoms. A random phrase disconnected from a user task merely creates noise. Before launch, deliberately replace content in a staging fixture, verify the alert, restore it and verify recovery.
How we made this guide
Four pages were served by a local Python HTTPServer and fetched with an HTTP client; the outcomes are in data.json. enterno behavior was checked against HttpChecker::checkKeyword() and the settings validator. The external-check model was compared with the Prometheus blackbox exporter guide. SerpAPI snapshots cover Google US and Yandex Moscow. Wordstat region 225 returned 49 for the main Russian phrase and 19 for a broader variant. Those counts are query volume, not broken-site prevalence.
FAQ
Why can I open the page while the keyword monitor says down?
Compare an external GET without cookies, its final URL and body. Your browser may render JavaScript content that a raw HTTP check does not see.
Can HEAD check a keyword?
No. HEAD has no response body. A content assertion requires GET.
Is searching for “Fatal error” enough?
No. A blank page or another placeholder will pass. Also require a normal-state marker.
What if copy changes often?
Provide a stable server-rendered marker or a dedicated health endpoint that reflects the actual feature.