Mobility apps and disabled riders — an audit of Uber, Lyft, Bolt, FreeNow, and DiDi

Five ride-hail and mobility platforms account for the overwhelming majority of app-summoned rides on three continents — and disabled riders use all five through interfaces, vehicle supplies, and driver-education programmes that diverge sharply by region and by operator. This dossier scores Uber, Lyft, Bolt, FreeNow, and DiDi on four pillars — app-level a11y, wheelchair-accessible-vehicle (WAV) supply and filter UX, service-animal handling, and regulatory-compliance posture — across twelve test cities, 440 test trips, and 120 hours of screen-reader recordings. The headline numbers are stark: median WAV pickup time across the sample sat at approx. 21 minutes against approx. 6 minutes for standard rides — a 3.5x penalty for needing the wheelchair-accessible filter at all. Composite a11y scores ranged from 62/100 at the top of the table to 38/100 at the bottom. Five years on from the DOJ-Uber service-animal consent decree, and ten months into the EAA’s transport-service scope under Article 4, the regulatory floor is being installed faster than the operational reality is catching up to it.

01 · How we audited the apps

Five operators, twelve cities, four pillars, four months. The audit ran from 6 January to 28 April 2026. Each operator was tested in the cities where it actually competes — Uber and Lyft in the North American sample, Bolt and FreeNow in the European sample, DiDi in the Latin American and Asia-Pacific sample, with overlap where multiple operators serve the same city. App-level a11y was tested on iOS 18.3 with VoiceOver and on Android 15 with TalkBack, against four observable surfaces: focus order through the booking flow, dynamic-type behaviour at 200% text scale, label completeness on interactive elements, and live-region announcements on driver-arrival and trip-state changes. WAV supply was measured by requesting wheelchair-accessible rides from a fixed grid of pickup points at three time-of-day windows and recording supply, wait time, and successful completion. Service-animal handling was tested with trained guide-dog teams requesting standard rides and logging refusals. Regulatory posture was scored by reading each operator’s published accessibility statement, driver education curriculum, and complaints-handling documentation.

The composite a11y score weights the four pillars at 30% app-level a11y, 30% WAV supply and filter UX, 25% service-animal handling, and 15% regulatory posture. The weighting reflects what disabled riders consistently say drives their actual experience of these apps: whether they can book the trip at all (app), whether a vehicle that fits them arrives (WAV), whether they get refused at the kerb (service animals), and whether the operator’s stated policy bears any relation to what the driver does (regulatory).

02 · The five-app a11y ranking

The composite ranking is closer than it looks. Lyft narrowly tops the table on the strength of its iOS VoiceOver build and a relatively mature WAV-filter UX in its North American markets. Uber follows a short distance behind — strong on driver-education coverage where the DOJ framework requires it, weaker on Android TalkBack focus order in newer feature surfaces. FreeNow ranks third on the European sample, with high marks on dynamic-type behaviour and lower marks on WAV supply. Bolt and DiDi sit at the bottom of the table, though for different reasons — Bolt’s app-level a11y is strong but its WAV programme is shallow; DiDi’s WAV supply varies by city, but its app builds diverge widely by region, with the LatAm Android build trailing the rest.

The five-app spread is narrower than the regional spread within any one operator. Where a rider lives, more than which app they open, predicts whether the trip works.

03 · WAV availability — twelve cities

Wheelchair-accessible-vehicle supply is the part of the audit where the gap between app design and operational reality is widest. Every operator in the sample exposes a WAV filter in its booking UI. The filter UX itself is fine — labels are reasonable, focus order works, the toggle is announced by both VoiceOver and TalkBack — but a filter is only useful if it returns a vehicle. In two of the twelve cities, the WAV-within-15-minutes rate exceeded 50%. In four of the cities, it sat below 20%. In the remaining six, it landed between 20% and 50%. The pattern is not random: cities with regulatory WAV-supply mandates layered on top of the ride-hail apps — New York’s TLC accessibility rules, London’s PHV licensing conditions — have measurably better WAV supply.

Two patterns deserve attention. First, the cities at the top of the chart are not the cities with the most progressive ride-hail operators — they are the cities with the strictest local regulatory mandates. London’s Private Hire Vehicle licensing regime requires operators to actively procure WAV supply; New York’s TLC rules tie ride-hail licensing to disability-access metrics. The operators respond to the mandate. Where the mandate is absent or soft, WAV supply does not arrive on its own. Second, WAV supply correlates more tightly with the local taxi industry’s accessible-fleet share than with the ride-hail operator’s published policy. Cities with mature accessible-taxi fleets that the ride-hail apps then dispatch through — London, New York, Toronto — outperform cities where the ride-hail operator is expected to source WAV supply from its general driver pool.

04 · Service-animal acceptance

Of the 165 guide-dog test calls placed across the twelve cities, approximately 51 — close to 31% — ended in a driver-side cancellation, a refusal at the kerb, or a no-show that the operator’s data later confirmed as driver-initiated. The headline rate is grim. The variation underneath it is worse: in two North American test cities the rate was under 18%; in three of the European test cities it landed between 28% and 35%; in two of the LatAm test cities it exceeded 45%. Driver education is part of the story, but only part. Where operators have tied disability-awareness training to onboarding under regulatory pressure — most prominently under the post-2021 DOJ framework — refusal rates are lower. Where the training is optional, refusal rates climb back toward historical baselines.

What disabled riders are asking for — and have been asking for since the National Federation of the Blind began coordinating service-animal complaints against ride-hail operators a decade ago — is straightforward: a documented zero-refusal policy, gated to onboarding, with consequences. The DOJ-Uber framework approximated this for one operator in one jurisdiction. Across the rest of the cohort, the policy text often exists; the enforcement does not.

05 · The DOJ-Uber settlement, five years on

The 2021 settlement in United States v. Uber Technologies, Inc. remains the single most consequential US enforcement action against a ride-hail operator on disability grounds. The matter focused on wait-time fees — Uber’s practice of charging riders for time spent boarding the vehicle, which the Department alleged disproportionately burdened riders with mobility disabilities who needed additional time. The settlement carried a $2.2 million payout, established a wait-time waiver framework, required policy and training changes, and installed a four-year monitoring window that has since been extended by amendment through 2027.

Five years on, three observations stand out. The wait-time-fee architecture has been restructured across the industry, not just at Uber — Lyft and several international operators followed with their own wait-time waiver programmes, partly to pre-empt parallel enforcement. Driver-education uptake on the disability modules, where the framework gates it to onboarding, sits in the high nineties, against an industry baseline closer to 14%. And the framework’s audit-and-reporting regime, while administratively heavy for the operator, has produced a reliable public-facing accountability stream that disability-rights organisations now reference in their negotiations with other operators.

What the settlement has not done is fix WAV supply or service-animal refusals at the operational floor. Both remain at rates that, if measured against the settlement’s text, would justify ongoing enforcement attention. The DOJ’s case-selection discipline — fewer than 200 federal website-accessibility actions in a decade, as Disability World reported in the DOJ enforcement tracker — means follow-on cases against ride-hail operators have been rare even where the conduct continues.

06 · EAA Article 4 and the transport-service scope

The European Accessibility Act — Directive 2019/882 — entered application on 28 June 2025 and brings a new layer of obligation to ride-hail operators serving European users. Article 4 lists the services that fall within scope. Among them: urban and suburban passenger-transport services and the websites, mobile applications, and ticketing interfaces that mediate them. The European Commission’s 2024 guidance treats ride-hail apps marketed in the EU as in scope.

Ten months of enforcement is too short a window to score the regime. What is visible is that seven member states have opened advisory actions against ride-hail operators in scope. One — Germany’s market-surveillance authority for digital services — has issued a formal compliance notice citing Article 4 against an operator (not publicly named at the time of writing). Several others have issued informal guidance letters. The published accessibility statements that EAA Article 4 requires are, by April 2026, present on the EU-facing pages of all five audited operators; their substance varies widely.

07 · Regional patchwork — what regulators are reaching for

Stepping back from the operator-by-operator scoring, the most striking pattern in the audit is the regional patchwork. North America runs on Uber and Lyft, layered on ADA Title III, a handful of state-level supplements, and city-level taxi-and-limousine commission rules where they exist. Europe runs on Bolt and FreeNow with Uber’s overlay in many capitals, layered on the EAA and on national equality laws like the UK Equality Act and Germany’s BFG/BITV. LatAm runs on Uber and DiDi with limited regulatory floor. APAC runs on DiDi, Grab, and Uber with national-law variation that ranges from Japan’s well-developed disability framework to jurisdictions where ride-hail is barely regulated at all.

Three regulatory threads are tightening simultaneously. The first is supply-side mandates of the New York / London type — accessibility metrics tied to operating licences. These work where they are adopted but require a regulatory architecture that most cities do not have. The second is conduct-based enforcement of the DOJ-Uber type — settling discrete claims and using the consent decree to install operational guardrails. These work where the enforcement authority chooses to use them. The third is the EAA’s accessibility-statement and structural-requirement architecture — a horizontal floor across all in-scope services. This works in the sense that the floor exists; whether the enforcement behind it is real will be the story of the next two to three years.

The supply mandate, the consent decree, and the horizontal floor all work — they just do not work in the same place, against the same operator, on behalf of the same rider.

For disabled riders, the practical upshot is that the choice of app matters less than the choice of city. A wheelchair user in London or New York moving by ride-hail has a measurably different experience from the same rider in Madrid or Sao Paulo, even when the app installed on their phone is identical. The audit’s scoring exercise — useful as a within-operator comparison and a within-region comparison — should not be read as a cross-regional verdict. The five operators are not running on the same regulatory floor; they are running on five overlapping floors that diverge by jurisdiction.

08 · The through line

Five years on from the DOJ-Uber settlement, ten months into the EAA’s transport-service scope, and a decade into systematic disability-rights organising against ride-hail operators, the audit’s findings are simultaneously encouraging and sobering. The app-level a11y work has moved measurably — the WAV filter, the booking flow, the screen-reader experience are all materially better than they were five years ago. The operational reality — whether a wheelchair-accessible vehicle actually arrives, whether a driver actually picks up the rider with the guide dog — has improved much less.

What regulators are reaching for is a tighter coupling between the app surface and the operational floor. London and New York show that supply-side mandates can deliver. The DOJ-Uber settlement shows that conduct-based enforcement can deliver on driver behaviour. The EAA’s horizontal floor across in-scope services in 30 member-state markets is the most ambitious of the three, and the one whose enforcement record is still being written. Disability World will continue to read it as the record arrives — in the EAA first-year report, in the DOJ enforcement tracker, and in the next ride-hail audit, scheduled for early 2027.