Your first scan will return hundreds of findings and a score around fifty. This is normal and it is not a judgement on your site. It is what happens when 41 checks run against markup nobody wrote with them in mind.
Most of that number is a handful of mistakes repeated across every page. The order you work in decides whether this takes an afternoon or a fortnight.
Before you start: do not sort by page
The instinct is to open page one and work down. Do not.
Sort by what comes up most. The top of that list is almost always a theme fault appearing on every page, and one change clears all of it. Working page by page means fixing the same mistake forty times.
The order
1. Site-wide fixes (30 minutes, clears the most)
Switches that supply what a theme leaves out, applied everywhere at once: a skip link, a visible focus outline, the page language, underlined links in body text, labels on core search and comment fields.
Read what each one might disturb before switching it on — some change behaviour rather than only appearance. The full list with caveats is in Site-wide fixes.
Expect a large drop in the count for very little work.
2. Contrast (an hour, usually a handful of values)
Almost every site fails contrast on two or three colours: body text grey, a muted caption, and a button state.
Fix the values, not the instances. One change to a colour clears every use of it. See colour contrast.
3. Alt text (the biggest in volume, fast in bulk)
Usually the largest single category. Use a bulk editor rather than opening media items one at a time, and write for purpose rather than appearance. See how to write alt text.
4. Theme and template faults (high leverage)
Headings emitted by the header, footer or archive templates. A second h1. A
missing main landmark.
Group findings by the markup behind them so a fault your theme repeats is judged once rather than once per page.
5. Forms (highest value per fix)
Fewer findings than the categories above, and far more consequence. A form nobody can complete is a transaction nobody makes. See accessible forms.
6. What is left, by severity
Critical and serious first. By this point there will be far less than the first report suggested.
7. The manual pass
Twenty minutes with a keyboard, then an hour with a screen reader. This is where you find the things no tool reported. See testing with a keyboard.
8. Write it down
An accessibility statement naming what is still outstanding, with dates. See how to write one that is true.
Reading the numbers while you work
Three figures matter more than the score:
Barriers found — settled by a check. These are wrong.
Needs a person to look — the check could not settle it. These may be fine.
Fixed — what you have cleared.
The score itself counts only what automated checks can settle, which is part of WCAG rather than all of it. A rising score means real progress. A perfect score means the automated part is done. More in Reading the report.
Two things that will confuse you
The browser pass. Contrast and target size need a rendered page, so a page that has only had the server scan has only had 36 of the 41 checks. If the score and the list seem to disagree, let both passes finish.
Counts dropping after an update. If totals fall sharply, nothing has been hidden. Earlier versions counted every scan of a page separately; only the most recent scan of each page is counted now.
How long this actually takes
On a small business site, roughly:
| Step | Time |
|---|---|
| Site-wide fixes | 30 minutes |
| Contrast | 1 hour |
| Alt text | 1 to 3 hours, depending on library size |
| Theme faults | 2 to 4 hours |
| Forms | 1 to 2 hours |
| Keyboard pass | 20 minutes |
| Screen reader pass | 1 hour |
| Statement | 1 hour |
A working week, spread out, for a site that has never been touched. Less than most people fear, and far less than it looks on the first report.
Questions people ask
Should I scan everything or one page first?
One page first. It finishes quickly and shows you the shape of the output before you commit.
My score went down after I fixed things. Why?
Most likely more pages were scanned, bringing in pages that had never been checked. The score is an average across scanned pages.
Do I need to fix everything before launch?
No, and holding a launch for a perfect score is usually the wrong trade. Fix the barriers, document what is outstanding, and keep going.
To start, the plugin is free on WordPress.org, and your first scan walks through it.