Reading your report

A report scores six dimensions from 0 to 100: performance, SEO & discoverability, design & trust, UX, content, and accessibility. Each is a card with its own score, a short summary, and a list of issues we found.

Above the dimensions sits a single overall score: a weighted roll-up that gives you one number to track over time. The dimension cards tell you where that number comes from.

Performance is anchored on a Lighthouse lab run for each page, fetched through Google's PageSpeed Insights API on both mobile and desktop (see Lab data). The other five dimensions come from an analysis pass over the pages we gathered: structure, metadata, copy, trust signals, and the lab metrics all feed the model that writes the summaries and scores.

Because the audit covers multiple pages, the scores describe the site, not one URL. Adding or removing pages from the crawl can move them. See Multi-page audits & sitewide scores.

Each dimension lists issues. The most actionable of these become action items you can track on the Actions board, with a status (open, in progress, complete, or ignored) and a priority (low, medium, high).

You change status and priority yourself; PageReflect never closes an item for you. Each item can also carry an assignee, a due date, and a comment thread, and you can request a written brief for an item or generate a handoff prompt for a set of items to paste into a coding agent. The board is your punch list for turning a report into shipped fixes.

The same board is reachable from the MCP server and the CLI, so an agent session can read, update, and comment on items without opening the app. See Connect PageReflect to Claude with MCP and The pagereflect CLI.

We capture a screenshot of each page we audit and show it alongside the findings, so you can see what the analysis saw. If a page failed to render, the report notes it rather than scoring a blank.

A report reflects the site at the moment of the run, measured in our cloud. It is a fast, repeatable signal, not a guarantee of what every visitor experiences. Pair it with real-user data when you need a population view: see Field data.