When someone asks what changed on a website, the best evidence depends on the question. A screenshot can show a moved button. Saved HTML can show a changed link destination. Preserved CSS can help explain why the same content now appears differently. Treating any one format as a complete record leaves gaps.
Build an evidence package around the decisions it needs to support. Keep the capture context alongside the files, and state the limits of any replay. A saved page is an observation under particular conditions, not a perfect copy of every possible interaction with the original site.
Use a screenshot to answer appearance questions
A screenshot records the pixels captured at a particular viewport and moment. It is useful for reviewing layout, visible text, clipping, overlays, and differences between desktop and mobile presentation.
It does not show every fact that affects the page. A link can keep the same visible label while its destination changes. A control can lose an accessible name without changing its appearance. Content below the captured region may be absent from a partial screenshot.
Record the viewport and whether the image covers the visible window or the full page. Keep the timestamp and page reference attached. Without those details, two images can look comparable while representing different contexts or different parts of the page.
Use HTML to inspect structure and content
Saved HTML can help investigate headings, links, form attributes, labels, and the content included in the captured document. It can also support a more precise comparison than an image when a small textual change matters.
Clarify what was saved. The original server response and the browser's rendered document can differ after JavaScript runs. A record that says only “HTML” may leave the reviewer unsure whether they are seeing the initial shell or the state after rendering.
HTML alone does not establish that an interaction worked. A button's markup may look plausible while an event handler is broken. Nor does the presence of an element prove that it was visible or usable under the observed conditions. For behavior, add test notes or other suitable evidence.
Preserve styles when presentation is part of the question
CSS helps explain how structure became a visual interface: spacing, positioning, colors, text treatment, responsive rules, and visibility. If the page's text is unchanged but its layout moved, the styles are an important part of the investigation.
A replay that retrieves current external styles can drift away from the original capture. The same saved HTML may then appear different months later. Record which styles and other assets were preserved, which remain external, and which could not be captured.
Even preserved styles do not freeze the entire rendering environment. Fonts, browser behavior, viewport size, images, and operating system rendering can influence the result. Playwright documents environment differences in screenshot comparison. Keep the captured screenshot as the record of observed appearance rather than assuming every replay will reproduce identical pixels.
Match evidence to the investigation
| Question | Most useful starting evidence | What else may be needed? |
|---|---|---|
| Did the deadline text change? | Captured text or rendered HTML | Screenshot and nearby context |
| Why did the call to action move? | Before-and-after screenshots | Preserved styles, structure, and viewport |
| Did a link point somewhere different? | HTML attribute comparison | Destination verification at the relevant time |
| Was an error available to a screen reader? | Recorded interaction and accessibility test notes | Relevant markup and application state |
| What did the visitor see during capture? | Screenshot | Session, consent, language, and loading context |
The table is a starting point, not a ranking of formats. A strong investigation often combines several of them. For example, an unchanged screenshot alongside a changed link destination can be exactly the evidence needed to explain a subtle but important update.
Keep a provenance record with the files
At minimum, preserve the requested address, final address after redirects, capture time with timezone, relevant environment, and capture outcome. Add the release or monitoring run reference when available.
Record failed or partial results honestly. Missing styles, an unavailable image, or a timeout can affect interpretation. Do not rename a partial capture as a complete archive simply because it produced an HTML file.
Checksums can help detect whether a saved file has changed since the checksum was recorded. They do not independently prove the truth of the page, its authorship, or the completeness of the capture. Keep those claims separate from routine file integrity checks.
For an audit package, the evidence handoff guide explains how to connect these records to stable findings and retest entries. Store only the material needed for the task, with the access and retention arrangements the organization has actually agreed.
Treat replay as an inspection aid
A preserved replay is useful when a reviewer needs to inspect the captured page without revisiting the original. It may deliberately disable scripts, navigation, forms, or external requests. Those restrictions can make the replay safer and more stable while limiting interactive behavior.
Label those limits clearly. A form that cannot submit inside an inert replay does not establish that the original form was broken. A current live page also cannot replace historical evidence when the question concerns an earlier capture.
OnChange's snapshot comparison combines preserved page evidence for review. The public scoping sample provides a concrete example with screenshots, archived page content, and downloadable report data. Inspect the formats together and use each one for the question it can actually answer.