Arabic TTS for Government Digital Services: How Natural Voice Closed an Accessibility Gap

A GCC government authority improved access to digital services with clear Arabic voice guidance, helping more citizens complete applications independently while reducing confusion, drop-offs, and support requests.

The Problem: Arabic Audio That Nobody Used

Government digital platforms across the GCC have come a long way. Residents can now handle a wide range of public services online, permit applications, utility management, and more. But most of these platforms were built around one assumption: that users are comfortable reading and navigating text-heavy interfaces. For many residents, that assumption does not hold.

One government digital services team noticed a clear pattern. Completion rates for complex, multi-step transactions were dropping among older residents and users whose primary relationship with Arabic is spoken, not written. Exit surveys and support call data kept pointing to the same thing: the interfaces worked, but they weren't accessible. Users wanted to hear instructions and confirmations, not read them.

The team had already deployed an English TTS voice that performed well. Arabic was a different story. The available Arabic TTS was mechanical, produced stilted phrasing on longer sentences, and used Modern Standard Arabic (MSA) pronunciation patterns that felt distant from how residents actually speak. User feedback was negative enough that the feature had been quietly switched off.

What They Needed

The platform team needed Arabic TTS that could do two things well:

• Narrate interface instructions for users who prefer audio guidance while navigating the platform
• Read back confirmation text, application summaries, submitted values, and reference numbers, in a natural, clear voice

The voice had to sound like a competent, professional Arabic speaker. This was a credibility requirement as much as an accessibility one. A government platform with poor Arabic audio doesn't just create friction; it signals that Arabic-speaking users are an afterthought.

On the technical side, the integration had to work at the component level, narrating specific UI elements on demand without requiring a major rebuild of the existing front end.

The Approach

CNTXT AI integrated Faseeh, Munsit's Arabic TTS model, into the platform's web interface via the Munsit API. Audio is generated on demand when a user triggers the audio guidance function on a given screen.

Faseeh was deployed across three specific use cases:

• Step-by-step narration: With audio guidance enabled, each on-screen instruction is read out in sequence with natural pacing between steps
• Form confirmation readback: Before submission, the platform reads key entered values back to the user in clear Arabic, giving them a chance to catch errors
• Status and reference number screens: Completion screens are narrated in full, which proved especially useful for users completing transactions on mobile in environments where reading a screen was impractical

Before full deployment, the platform team tested Faseeh with a focus group across several age groups. The feedback was noticeably different from every previous Arabic TTS experience they had tested.

Results

Service completion rates for the platform's most complex multi-step transactions improved among users who engaged with audio guidance. Drop-off at confirmation screens fell; the team attributed this in large part to users verifying their submissions by listening, rather than re-reading dense text.

The support center saw a measurable drop in calls from users who had submitted forms with errors they hadn't caught. The readback function was cited repeatedly in support call transcripts as something users relied on to check their entries before hitting submit.

Since launch, the team has extended the Faseeh integration to additional service categories. Work is ongoing to add Arabic audio to outbound notification output, a use case that requires asynchronous TTS generation rather than real-time response.