Read before-and-after content, HTML, and JSON differences

Interpret highlighted evidence for code, content, headers, URLs, and structured data.

Last checked 20 August 2026

TechDash formats incident evidence according to what the check measures. Change checks usually compare Before with After. The dashboard card can relabel After as Now, while the detail dialog retains the check's full comparison labels.

Read highlighting

When both sides are available, TechDash compares the displayed text:

  • Removed text is highlighted on the Before side.
  • Added text is highlighted on the After side.
  • Unchanged text remains unhighlighted.

Multiline values are compared line by line. Single-line values are compared in word and whitespace tokens. The algorithm is intended to make the changed region readable. It is not a source-control diff and does not attempt to infer the author's edit.

Understand displayed formatting

TechDash can normalize evidence before display:

  • JSON and JSON-LD are pretty-printed when they parse successfully.
  • Markup-like values and selected code-oriented checks use a scrollable monospaced block.
  • Canonical URL comparisons that point to the same normalized destination can be represented as canonical link tags.
  • Robots and X-Robots-Tag evidence can be reconstructed as the relevant meta tag or header line.
  • Numeric values and word counts are formatted for readability.
  • Analytics identifiers are normalized and summarized.
  • Linked-resource failures can show URLs with HTTP status labels rather than raw storage text.

This means the comparison is the check's interpreted evidence. It is not guaranteed to be a byte-for-byte comparison of two downloaded files.

Body content and saved HTML

For body-content changes, the compact evidence normally comes from visible body text captured by TechDash, not scripts and styles. When the check stores a larger HTML comparison, the database result can contain a local reference rather than the whole document. The dialog loads the saved HTML from local compressed storage and displays Loading saved HTML while resolving it.

The detail view then compares the saved server-rendered or browser-rendered HTML associated with that result. If a referenced local file is unavailable, the evidence may be empty or shown as not recorded. This does not cause TechDash to refetch historical HTML from the website.

Comparisons that use other labels

Not every two-column view means historical before and after:

  • Current-state checks use Best Practice and Current.
  • Limit checks can use Best Practice and Current Amount.
  • Main-content JavaScript divergence uses Server Side Rendered HTML and Client Side Rendered HTML, with word counts rather than complete page source.
  • Some checks use labels such as Previous Word Count and Current Word Count.

Read the labels before deciding that a highlighted difference is a regression.

Evidence that is not shown as a diff

TechDash suppresses generic comparisons for list-style incidents, fingerprints, and evidence that is better explained structurally. These incidents use URL cards, grouped values, or tables. Examples include duplicate titles, hreflang target findings, Search Console query groups, high-value anchor-text changes, and redirect mappings.

The dashboard also clips short evidence at a fixed display limit and does not show stored HTML references or fingerprint-only values. Open the incident for the full supported presentation.

Interpret a difference safely

  1. Read the check description and evidence labels.
  2. Identify the exact added, removed, or changed signal.
  3. Compare the check time with deployments or CMS edits.
  4. Open the live URL to confirm the current state.
  5. Decide whether the signal is intentional.
  6. Fix the source and select Check & Verify, or use the appropriate acceptance action.

Formatting can expose semantically equivalent text as changed if the monitored representation differs. Conversely, normalization can hide irrelevant formatting differences. Judge the SEO or operational effect of the monitored signal, not the color alone.

Related articles