Website Monitoring

Website monitoring (screenshot-based) is the practice of tracking visual changes to a website over time by capturing screenshots at regular intervals and comparing them to detect differences.

Related workflow terms: Screenshot Automation, Batch Screenshot.

Screenshot-based vs uptime monitoring

Uptime monitoring answers a simple question: is the server responding? It sends a request and checks for a successful HTTP status code. If the server returns a 200 OK, the site is considered "up." But a 200 response says nothing about what the page looks like. A page can load successfully while displaying broken layouts, missing images, wrong content, or an error message embedded in the body.

Screenshot-based monitoring goes deeper. It captures what the page actually renders — the visual output a real user would see. By comparing captures over time, it detects changes that uptime checks cannot: layout shifts, content edits, style regressions, and broken visual elements.

The two approaches are complementary. Uptime monitoring catches server-level outages. Screenshot monitoring catches visual-level regressions and content changes that happen while the server is running normally.

Where screenshot-based monitoring is used

• Competitor tracking — capturing competitor pricing pages, feature lists, and landing pages at regular intervals to detect changes.
• Brand and compliance — monitoring your own public-facing pages to ensure content, disclaimers, and branding remain correct and up to date.
• QA and staging — capturing staging environments after each deployment to verify that the visual output matches expectations before promoting to production.
• SEO monitoring — detecting changes to page structure, headings, and content that could affect search rankings.
• Regulatory pages — tracking legal pages, terms of service, and privacy policies for unauthorized or unintended changes.

How it works

A screenshot monitoring workflow has three stages: capture, compare, and alert.

In the capture stage, an automated tool — typically a headless browser — navigates to the target URL at a scheduled interval and saves a screenshot. The capture should use a consistent viewport size, device pixel ratio, and wait strategy so that differences between captures reflect actual page changes, not variation in the capture setup.

In the compare stage, the new screenshot is compared against a previous baseline. The comparison may use pixel-level diffing, perceptual diffing, or structural analysis. Regions with dynamic content — ads, timestamps, live counters — can be masked to prevent false alerts.

In the alert stage, if the comparison exceeds a defined threshold, a notification is sent — via email, Slack, webhook, or another channel. The alert typically includes the before and after screenshots and a visual diff highlighting the changed areas.

Automated capture tools can schedule website screenshots at intervals for visual change tracking, running the entire capture-compare-alert loop without manual intervention.

For a Shotomatic-style workflow, the sweet spot is usually a fixed capture template plus a small set of masked regions. That keeps alerts focused on meaningful page changes instead of noise from rotating content or timestamps.

Common mistakes

• Not masking dynamic regions. Ads, rotating banners, timestamps, and user-specific content change between every capture. Without masking, every comparison triggers a false alert, and real changes get lost in the noise.
• Using inconsistent capture settings. Varying the viewport size, device pixel ratio, or wait timing between runs introduces differences that are not real changes. Standardize every parameter and run captures in the same environment.
• Monitoring too infrequently. Capturing once a day may miss changes that happen and revert within hours. Match the capture frequency to how quickly changes matter for your use case.
• Ignoring authenticated pages. Many important pages — dashboards, account settings, pricing for logged-in users — are behind a login. If the monitoring tool does not authenticate, it only captures the login screen.