Following the latest changes on the PBIX A11y website (catch up at PBIX A11y update: alt text for groups, a rebuilt tab order check, and a clutter fix), I’ve brought the same fixes over to the PBIR A11y CLI.
First of all, I have added a new check. The CLI now compares your authored tab order (the sequence you set in the Selection pane) with the order a sighted user would naturally read the page in. That expected order comes from each visual’s x and y position on the page: visuals are grouped into rows by how much they overlap vertically, then read left to right within each row, so a button sitting in the top right still counts as the same row as a page title on the left even if the button is a little higher or lower. If your authored order doesn’t match this, it tells you exactly where they diverge. This check is advisory, not a hard fail, because only you know the reading order you intended. Still, it’s a useful reminder for tab orders that were set once, early on, and never revisited as the page grew.
If your reports use Power BI’s visual groups, there’s more. A visual sitting inside a group has its position measured relative to that group, not the page, so anything working out layout or overlap was reading the wrong numbers the moment groups were involved. That’s fixed: the CLI now works out a visual’s real position by walking up through however many groups it’s nested inside, so clutter and overlap checks see accurate coordinates regardless of grouping. Duplicate tab order detection is scoped correctly too, comparing a visual only against its actual siblings rather than every visual on the page, so a value shared between a top-level visual and one inside a group no longer gets flagged as a clash it was never going to cause.
Alt text for groups was addressed a little earlier, thanks to a contribution from John Kerski: groups now get the same missing, too-short, and placeholder alt text checks as any other visual, with guidance that steers you toward describing what the group holds rather than counting items, since a count goes stale the moment someone adds or removes a visual.
If you’ve run pbir-a11y check before, it’s worth running it again. Every report may see the new tab order check for the first time, and if you use groups, your clutter and duplicate tab order results specifically may look different too, more accurately so.
One limitation worth being upfront about: nothing here can truly check for a “meaningful” sequence in the strict WCAG 2.4.3 sense. The layout-inferred order is a spatial heuristic, top to bottom then left to right, and it has no way of knowing that your report’s story deliberately opens with a KPI card even though a chart sits above it, or that a filter panel belongs at the end on purpose. That judgment stays yours to make, which is exactly why this check is advisory rather than a fail. It’s not a gap on the roadmap either: no heuristic based on position can tell you what a sequence actually means, only a person reading the report can.



