Conformance status
WheelsUp is partially conformant with WCAG 2.1 Level AA. "Partially conformant" means most of the standard is met, but not all of it has been verified.
We are not claiming full conformance, for one honest reason above all others: we have not tested WheelsUp with a screen reader. Automated measurement and keyboard testing can show that a name exists; they cannot show that it reads sensibly. Until we have done that testing, a full conformance claim would be a claim about work we have not done.
This statement covers militarywheelsup.com (the marketing site, guides and legal pages) and app.militarywheelsup.com (the planner).
What we measured
An accessibility audit on 2026-08-04 produced ten findings. All ten have been addressed, and a second measurement pass the same day re-checked them and went looking for what the first pass missed. Specifically:
- Colour contrast. 2,632 text-and-background pairs measured across 18 pages, with every colour resolved from what the browser actually paints — including text sitting on gradients and see-through panels, where the effective background was computed layer by layer rather than guessed at. Text now meets the 4.5:1 minimum (3:1 for large text) everywhere we measured.
- Keyboard. Every interactive control is reachable and operable by keyboard. We tabbed through the site with a real keyboard rather than simulating it in script — 25 presses, 25 different destinations, nothing trapped.
- Reflow and zoom. No horizontal scrolling at 320 px, or at the equivalent of 200% browser zoom.
- Text spacing. Applying the standard's own spacing overrides (line height, letter, word and paragraph spacing) clips or overlaps nothing.
- Structure. Every page has one
<h1>, no skipped heading levels, a<main>landmark, named navigation regions, and a "skip to main content" link as the first thing you reach. - Status messages. The planner's calculators announce their result when it changes, so a screen-reader user is told the numbers they asked for have arrived.
- Moving content. The scrolling list of tools on the home page has a pause button, stops entirely if your system asks for reduced motion, and pauses on hover.
- Images and graphics. Every image has alternative text. The transition timeline — which is drawn as a chart — also publishes every event, with its full name and real dates, as a table for screen readers.
What we have not tested
- Assistive technology. Nothing has been tested with NVDA, JAWS, VoiceOver, Dragon, or a switch device. This is the largest gap on this page and the main reason our claim is partial.
- Independent audit. No third-party accessibility audit has been carried out. Everything here is our own measurement.
- Users with disabilities. We have not run usability testing with disabled users. Conformance is a floor, not a substitute for that.
- Cognitive and low-vision testing beyond the mechanical checks described above.
Not applicable: WheelsUp contains no audio or video, so the captions and audio-description criteria (WCAG 1.2.x) do not apply.
Known gaps
- Screen-reader wording is unverified. Controls, the timeline table and the calculator announcements have programmatic names, but we have not heard them read aloud. Some will be clumsier than they should be.
- Contrast on layered backgrounds is computed, not sampled. Where text sits on a gradient we tested against the worst-case point of that gradient. This is a conservative method, but it is arithmetic rather than a photograph of the screen. It now includes text that is partly see-through and text on the planner's gradient-drawn chart bars — two cases it got wrong until 2026-09-28, described under "How we test" below.
- PDF and CSV exports have not been assessed for accessibility.
- Third-party components. Sign-in uses Cloudflare Turnstile and payments use Stripe Checkout; those interfaces are theirs, and we have not audited them.
If you rely on any of these and hit a wall, please contact us — a specific report from a real user moves an item up this list faster than anything else.
What should work
- You can reach every control with Tab, and the focused item is always visibly outlined.
- A "skip to main content" link is the first thing you reach on every page, so you don't have to tab through the navigation on every view.
- You can zoom to 200%, or use a 320 px-wide screen, without the page scrolling sideways.
- Sign-in, onboarding and the calculators report errors in text, mark the specific field that needs fixing, and move focus to it.
- When a dialog opens, the rest of the page is taken out of the tab order until you finish or close it.
- If your system asks for reduced motion, animations stop.
- The site follows your operating system's light or dark preference.
Tell us about a barrier
We want to hear about it — especially the things this page does not list. Our user base is, by definition, more likely than most to include people with disabilities: we ask you for your VA rating.
Contact us about an accessibility barrier, and please include the page you were on, what you were trying to do, and the browser and any assistive technology you were using.
We aim to respond within five business days. WheelsUp is run by one person, so a fix may take longer than a reply — we will tell you which it is and when to expect the other.
How we test
Assessment was carried out by the WheelsUp team (self-assessment) on 2026-08-04, using three deliberately different methods, because each one is blind to something the others catch:
- Source inspection across the 31 marketing pages published at that date, and the planner, for the failures that are decidable from markup — missing alternative text, unlabelled controls, heading order, language attributes.
- Live measurement in a browser for anything that depends on rendering: contrast against the actually-painted background, layout at 320 px and at 200% zoom, text-spacing overrides, landmark structure.
- Real keyboard input for everything focus-related. We learned this the hard way: focus behaviour driven by script does not always match what happens when a person presses a key, and testing it the wrong way produced a confidently wrong result in both directions before we caught it.
The checks are pinned by automated tests that run before every release, so the things we fixed cannot quietly regress.
Those tests run against every page we publish, not only the ones that pass measured. The marketing surface has grown from 31 pages to 57 pages since 2026-08-04. On 2026-08-24 we re-ran the second and third methods above — live measurement in a browser, and real keyboard input — across all 37 pages that existed then, and on 2026-09-17, when twenty guides were added, across all 57, so no page here is published on assurance inherited from a page that was measured instead of it. The August re-run was 18,068 text-and-background pairs measured in both light and dark, none of them failing, and 1,673 real Tab presses: every page tabbed from top to bottom by a browser receiving actual key events, not a script calling focus(), because the two do not behave the same and only one of them is what a person experiences. The September re-run found no contrast failure on any of the 57 pages in either theme and reached 2,711 focus stops by real Tab presses. It found one problem, on the home page: with the text-spacing overrides WCAG 1.4.12 describes, the hero's headline pushed the page sideways at a 1,280 px window. It was fixed the same day, and the home page passes that check now. Both methods were run again on 2026-09-25, because the pages themselves had changed since September: every page on this site received a new navigation bar and footer, the stylesheet behind 56 of them was rewritten, and the section headings on the policy pages were rebuilt. A page that changes after it was measured is a page published on assurance that no longer describes it, which is the thing the sentence above refuses to do. That re-run measured 47,911 text-and-background pairs across the whole surface in both light and dark, none failing, and 3,102 focus stops by real Tab presses across all 57 pages, with no problem found on any of them. It also caught a defect this page had not: the topic illustration on four policy pages had come loose from its heading and was overlapping the summary box beneath it. That is fixed.
That re-run found something, and it is worth naming rather than quietly fixing. Our contact dialog told assistive technology it was modal while leaving the page behind it in the tab order — so a screen-reader user had the rest of the page correctly hidden, and a keyboard user could press Tab three times and end up on a link underneath the dialog they thought they were in. It is fixed, and the test that is supposed to hold us to the sentence above now checks every dialog we ship, on the site and in the planner, instead of the one place the fix already happened to be.
The next thing we have to name is a fault in our own measurement. On 2026-09-28 we made the contrast check stricter, and it found five pieces of text below the minimum that our earlier runs had passed. The check had treated partly see-through text as if it were solid, and it could not read the colours of the bars in the planner's timeline chart, which are drawn as gradients. Measured correctly, the timeline labelled every bar in white whatever the bar's colour, and on the paler bars that fell as low as about 1.5:1 against a minimum of 4.5:1. Two faded notes in the planner, in TSP Growth and Lifetime Monte Carlo, measured 3.75:1 and 3.58:1. Two labels in the illustration on our home page measured 3.96:1 and 4.00:1, because they sit on a tinted row and had been measured against the panel beneath it. So the "none failing" figures above were wrong in those five places. All five were fixed on 2026-09-28 and live by 2026-09-29, and the check now accounts for transparency and for chart gradients. We then ran both methods again on 2026-10-02 with the stricter check: 48,342 text-and-background pairs across the whole surface, and 26,641 more on the planner's public demo, in both light and dark, none failing, and 3,102 focus stops by real Tab presses across all 57 pages, with no problem found on any of them.
Standard applied: Web Content Accessibility Guidelines 2.1, Level AA.