An accessibility audit tells you what was true on the day it ran. The site does not hold still after that.
The day after remediation is the best shape the site will ever be in. Somebody tested it, wrote up what was broken, and the broken parts got fixed. The templates are clean, the color contrast passes, the headings run in order. Then everyone goes back to work, and work means publishing.
Nobody sets out to undo it. A staff member uploads the event flyer as a PDF, and the PDF started life as a scan of a printed page, so there is no text in it at all. A volunteer bolds a line and bumps the size instead of using a heading, because the real heading looked too big next to the photo. A marketing person writes “click here” three times in one paragraph because the sentence read better that way. Each of those decisions is small, defensible, and made by someone who was never told otherwise.
I do this to my own site. One of my notes posts had a 2001 screenshot carrying a 326-character alt attribute, and I wrote that alt myself. The description was accurate. It was also long enough that nobody listening to it would sit through the whole thing to find out what the image was. My own tool caught it weeks later, when I audited this very site with it. The fix was a short identifying alt plus the full description in a visible caption, where sighted readers could use it too. I work in this material every week and I still did it, because I was writing a post and a rich description felt useful in the moment.
The scan you stop running.
An organization finishes an audit, fixes what it found, and sets up a quarterly re-scan because that felt like the responsible cadence. The first re-scan comes back with a list nobody expected, and almost all of it came from content published in the three months since the audit. The second list is longer. By the third quarter, running the scan means admitting how far behind the content has drifted, and scheduling it keeps sliding to next month.
People do not stop running the scan because they stopped caring. They stop because the result got too big to act on, and putting it on the calendar starts to mean setting aside a bad week. A recurring check holds its place on the calendar when each run is small enough to work through in an afternoon, and that only happens when not much broke in between.
That is the part of accessibility work that has nothing to do with a scanner. No tool gets you there, including mine.
What keeps the number small.
Two pieces do the work, and they are separate purchases for a reason.
The first is a written procedure. Not a policy, and not a statement on a page about commitment, but the operational version: who checks what before a page goes live, what the check consists of, who approves it, who owns the fix when the check fails, and how often the whole site gets looked at again. Public entities are already used to this shape, because they publish grievance procedures and records-request procedures and know what a real procedure reads like. A written procedure survives the person who wrote it leaving, which a habit does not.
The second is the people who publish knowing what to look at. That is training, and the useful version happens on the client’s own CMS with the client’s own content, not on slides. Alt text, including when to leave it empty. Headings as structure rather than as sizes. Descriptive links. What to do with a Word document somebody pasted in. It takes a couple of hours and it changes what shows up in the next scan more than any amount of remediation does, because remediation fixes what already happened and this changes what happens next.
Neither piece certifies anything. Training does not make a site conformant and a procedure does not either. What they do is keep the quarterly number in the range where a person can look at it.
Where this is going.
I am building both of these into named offerings, which is a change from how I have worked so far. Documentation and training have always ridden along inside my builds and support work, and folding them in that way made them invisible; a client got them without ever deciding to buy them, and nobody can maintain what they did not know they were handed.
If you had an audit last year and have not looked since, the honest first step is finding out where the site sits today, and that is what the accessibility audit is for. If you already know the answer and the problem is that the number keeps growing, that is the other conversation, and I would rather have it before the third quarterly scan than after.