Baseline capture
When a customer adds a site, RankLatch requests its public homepage, checks the conventional sitemap locations, and stores one snapshot per selected page. A snapshot contains the final URL and response status, document title, meta description, canonical URL, robots meta and X-Robots-Tag directives, H1 text present in the HTML source, Open Graph and Twitter-card values, the declared favicon URL, structured-data block count, approximate source-body word count, document language, and response time.
The crawler is read-only. It does not authenticate to the content management system, execute writes, submit forms, or alter the website.
Exact-field comparison
Each scheduled run extracts the same normalized fields. RankLatch compares the fresh snapshot with the immediately preceding snapshot so an unchanged problem does not create the same alert every day. The original approved snapshot remains stored as restoration evidence.
Severity rules
Severity is rule-based and auditable. A page that gains noindex, becomes unavailable, loses its title, or points its canonical at another domain is critical. Title, heading, internal canonical, or removed homepage favicon declarations require review. Description, schema-count, social-preview changes, favicon replacements, and smaller copy changes stay in the dashboard evidence log; removal of a declared social image is elevated for review.
CriticalLikely to remove, redirect, or materially misrepresent an indexable page.
ReviewCould alter search intent or the approved result presentation.
NoticeUseful evidence, usually not worth an interruption on its own.
What is intentionally ignored
RankLatch does not diff the full DOM. Session identifiers, timestamps in markup, analytics tags, CSS class names, ad attributes, animations, and other implementation details are outside the alert model. Source-body text is represented only by an aggregate word-count threshold, so a sufficiently large rotating text block can still create a lower-priority event.
Known limits
RankLatch sees what an unauthenticated crawler receives. It cannot prove that Google has indexed a page, diagnose ranking loss, verify analytics attribution, inspect content available only after login, or detect a visual problem that leaves the monitored fields unchanged. JavaScript-only pages may expose less source content than a rendered browser.
A clean result is not a guarantee of rankings or complete technical health. It means the monitored public signals did not trip the documented checks at that time.