Reading the verdict

The check has finished and the rail has filled with a result. You want to know two things: how bad it is, and which parts of your draft it is talking about. The panel is built to be read top to bottom.

The headline

The first line is the verdict, and there are three of them:

• Needs attention — something serious was flagged.
• Worth a look — problems were found, but milder ones.
• Reads clean — nothing to fix was found.

Under it sits a counts line such as 4 issues · 2 to keep · 3 clear and a coloured bar showing how the findings break down by severity. Beside the headline is a small info button, How the problem map works, whose tooltip states the rule for the whole panel: Hover a problem to preview its highlight. Click to keep it focused.

A fourth headline, Clean in what was read, replaces Reads clean when the read did not cover your whole draft or lost part of itself. In that state everything below is scoped: a factor listed under Checked and clear is clear for the part that was actually read, not for the whole message. The panel says so outright, and a strip offers you a re-run.

Problems

Problems carries the count of what was flagged. In a sequence the rows are grouped by message step, so you can see which touch the trouble is in, and each group header carries its own count of annotations.

Each row shows the claim, a short description, the factor it came from with where in the draft it sits, and a severity badge. There are four, strongest first: Must fix, Should fix, Worth fixing, Optional.

Within a step, rows come strongest first, so what you meet at the top is the worst of that step and not the start of it. Step groups themselves stay in message order, which means the very first row of a sequence is step one's worst finding and not the read's. On a single message there are no steps, so the whole list is strongest first.

A read with more than twenty findings does not show them all. The panel keeps the twenty most severe, chosen across the whole draft, and the rest sit in a collapsed drawer labelled something like 7 more findings · the rest of this read. Two things are easy to miss about the drawer:

• The held-back findings are still highlighted in your draft even while it is shut. Clicking one of those highlights opens the drawer to reveal its row.
• Step headers then read 2 of 8 annotations, because the count on the left is what is listed and the count on the right is what exists.

Getting from a finding to the text

Every finding that could be tied to specific words paints them in the draft. Hovering a row in the panel washes the matching words; hovering the words raises the matching row. Nothing is committed by hovering. Clicking either side opens that finding's card and scrolls both into view.

Some findings have a row and no highlight. That happens when the quoted words no longer occur in the text, when they occur more than once and marking one of them would risk marking the wrong copy, or when the finding is about the message as a whole rather than a passage. Those rows still open their card, which appears beside the row. On a problem the card gives you the reason and what a fix has to satisfy, not the words themselves.

Only one tab's findings are painted at a time. Switching to Simulation swaps every mark in the draft for that reader's findings, and switching back swaps them again.

Worth keeping

Worth keeping lists the lines the check says are working, marked in the draft with a green dotted underline instead of a severity colour. Their rows carry a Keep tag where a problem carries a severity badge, and the note above them explains why they are worth reading: Keep these lines in a rewrite — they are the only findings that say what is working.

This section only appears when the check found something to keep.

Checked and clear

Checked and clear lists the factors that came back with nothing to flag. These have no highlights in the draft, for the obvious reason. Expand a row to see Looked for patterns like and what that factor was hunting for, which is the honest way to read a clear result: it tells you what was searched for and not found.

When the draft moves on

Edit anything after a check and the highlights fade to a dim wash while your words stay fully readable, with a Document changed strip pinned under the verdict. The findings are still true about the version that was checked. They no longer point at text that exists, that is all. Re-running is what re-anchors them.

Where to go next