Yalla Nihki: How a Chatbot Helped Hundreds Seek Mental Health Support in Jordan

In Jordan, a country that hosts the second largest number of refugees per capita, a significant portion of the population is affected by mental health challenges.

Following migration, COVID-19, and persistent socio-economic insecurity, some reports suggest that 95% of the people who need of mental health services have never received treatment. The InforMH program – known locally in Jordan as Yalla Nihki or Let’s Talk - uses behavioral science and human-centered design to bridge this gap.

By addressing the informational and behavioral barriers that prevent people from seeking mental health and psychosocial support, Yalla Nihki deploys encouraging outreach messages through digital and in-person platforms.

In 2024, we rolled out a pilot study of the latest iteration of the intervention.

Through Yalla Nihki, the IRC achieved promising results: the pilot study showed an 18% reduction in people scoring into our highest psychological distress category, and an 11x increase in help-seeking behavior – all for just $10 per client.

To date, over 1,400 people have actively participated in the program in Jordan, which has led to a few key findings:

57% of users sought mental health services

In a country where fewer than 5% of people in need of mental health care ever access services, the results from Yalla Nihki are striking. Among those who completed at least the first week of their journey, 57% went on to seek care — a rate 11 times the national baseline, and 28 times higher than comparable interventions, which have seen a help-seeking rate closer to 2%, for the same hotline.

This steep increase in help-seeking suggests that behaviorally informed, personalized interventions like Yalla Nihki can succeed where traditional awareness campaigns have fallen short: bridging the gap between intention and action in a meaningful and measurable way.

Making Care-Seeking Easier: The power of behavioral design

Behavioral science tells us that people are more likely to act when the process is simple and seamless. Our pilot confirmed this. We gave participants the option to call the service provider or receive a callback: 75% chose to be contacted by a service provider instead of making the call themselves.

By simply removing the burden of initiating contact from the participant (often the most psychologically difficult step) we significantly impacted uptake.

This insight reshapes how we think about referral systems. It’s not enough to provide a phone number; people are more likely to follow through when we help to carry the emotional weight of that first step.

80% of users reported a reduction in their distress levels

To understand the impact of the intervention on users’ mental health, we invited participants to complete self-assessment questionnaires before and after their chatbot journey. These assessments drew from three validated tools: the K10 scale for psychological distress, the WHO-5 for wellbeing, and the Social Distance Scale for stigma.

While over 400 people took the original assessment, not all users completed the follow-up survey; this is common in digital interventions. However, among the 66 participants who completed both pre- and post- assessments, the results were encouraging:

• 80% reported reduced levels of psychological distress
• 70% reported improvements in overall wellbeing
• Stigma-related attitudes remained stable

This finding could challenge a common assumption in mental health programming: that we must first change attitudes and beliefs to change behavior.

In our case, help-seeking and self-care behaviors increased significantly, even though stigma did not change significantly. In other words, while distress and wellbeing improved, attitudes toward mental health remained mostly the same. This highlights a well-documented behavioral pattern, that actions often change before beliefs.

By encouraging people to take small steps, such as trying self-care or speaking to a professional, programs like Yalla Nihki may help users update their attitudes over time, based on their own lived experiences. Behavior change can be the catalyst for mindset change, not just the result of it.

Yalla Nihki costs only $10 per user

With support from the IRC’s Best Use of Resources (BUR) team, we estimated the total cost of the intervention to be approximately $10 per user. This figure includes everything: social media outreach, chatbot operations, supervision by a mental health specialist for high-risk cases, and ongoing monitoring and coordination by the project team.

As we scale, we anticipate this cost will decrease even further. Deeper integration of AI, for example, to automate parts of data monitoring and response routing can reduce the need for manual oversight, freeing up staff capacity and lowering overhead costs.

The pilot also helped us understand which outreach strategies deliver the best return on investment. A six-week Facebook ad campaign reached hundreds of thousands of people and generated the majority of our 1,443 chatbot users for about $0.50 per user. In contrast, printed materials placed in multiple pharmacies across the country only generated 30 new users, at an acquisition cost of over $40 per person – an entry point we will not continue with.

Another promising and affordable strategy was community influencer outreach (asking influential community members to invite the people they think would benefit most), which achieved a 45% conversion rate at just $4.60 per user. This culturally grounded approach may offer an effective complement to digital ads, particularly in communities where trust or digital access are barriers.

Designing for Scale and Inclusion

The pilot phase of Yalla Nihki was not only a test of feasibility, it was an opportunity to experiment with different strategies and assess what actually drives impact. The pilot also gave us the opportunity to test new features aimed at expanding audiences the chatbot could serve. These included the ethical integration of AI and the development of a non-text-based journey, which we plan to build on in future iterations.

Leveraging AI for Scale and Personalization

To streamline operations and increase cost-effectiveness, we embraced the use of AI in Yalla Nihki to identify participants who might pose a risk to themselves or others. For example, we leveraged natural language processing (NLP) to train the chatbot with over 25 Arabic trigger words and phrases that might implicate self-harm. This enables the chatbot to flag a specific chat, where a user needs support, to our psychologist, who can then engage in a 1:1 chat with that person.

In the future, we aim to expand the use of AI and NLP, especially in response to users expressing a desire for more open-ended questions that allow them to better articulate their feelings. Together, generative AI and NLP can help improve our understanding of user responses and craft relevant replies, enabling more personalized and dynamic conversations for each individual.

Developing a version for people with low-literacy

Text-based digital tools often leave behind people with low literacy. But in Jordan, we found that many people with limited reading skills use WhatsApp through voice notes and emojis. Inspired by this, we created an audio-based version of the chatbot.

The result? A flexible, intuitive experience where all content was delivered via voice notes, and choices are represented by emojis. Initial testing with a smaller group of 30 users was overwhelmingly positive. Not only did low-literacy participants feel included (some expressed feeling seen for the first time), but literate users also appreciated the convenience of audio, especially busy caregivers and working professionals.

Navigating Roadblocks

While much of our discussion has focused on the opportunities and positive outcomes from the pilot, it was equally important for highlighting pain points. We identified several barriers limiting the platform’s ability to reach and support individuals in need of mental health services.

High drop-off at the consent stage and a redesigned onboarding

One of the biggest bottlenecks came right at the start: of the 1,443 users who initiated a chat, only 56.4% completed the consent process. Many dropped off after receiving a long and formal text they didn’t fully understand, before seeing the benefits of the program. .

We’ve since redesigned onboarding to feel more intuitive and human. Users now receive a warm welcome, brief behavioral nudges, and a preview of how the chatbot can help before being asked to consent. We also replaced the formal PDF with a new illustrated infographic to simplify the process. Early A/B tests suggest these changes reduced early drop-offs.

Limited service provider capacity, and plans to expand access

Yalla Nihki significantly boosted help-seeking, but some local providers were unable to keep up with increased demand. Many partners in our pilot locations are facing staff shortages and funding gaps.

To scale responsibly, we’re expanding our referral network, formalizing new partnerships, and building out self-guided support within the chatbot, including psychoeducation, habit-building tools, and audio-based resources.

Importantly, we’re also currently testing the integration of a digital treatment option: a single-session intervention (SSI) for users who don’t require intensive care. This would:

• Support people on the move or awaiting services
• Provide stopgap care for those on waitlists
• Reserve human support for users in highest distress

These tiered support options will allow us to reach more people, more efficiently, without compromising the quality or care people receive.

Next steps

The promise of Yalla Nihki remains undeniable: It is a low-cost, evidence-informed, culturally resonant tool that brings mental health support into people’s lives in a way that is personal, practical, and scalable.

However preventative digital-first mental health interventions are still underfunded, and at risk of being sidelined in an increasingly stretched humanitarian landscape.

Having addressed the challenges identified from the pilot, we are now preparing to launch a rigorous randomized controlled trial (RCT) to more precisely measure impact.

While our results are highly promising, an RCT will provide the causal evidence needed to advocate for broader shifts in mental health policy and programming.

We’re also moving forward with adaptation. We’ve already secured funding for, and are currently adapting, the intervention in Mexico where it will support migrants, asylum seekers and refugees. This work is underway, and offers valuable insights into how the model can flex across cultural and linguistic contexts.

At the same time, we are actively seeking funding and partnerships to replicate Yalla Nihki in other regions and tailor it to new populations, such as youth, displaced communities, or frontline workers. With modest investment, we could reach over 100,000 people in Jordan within the next 2 to 3 years.

With the right partnerships, the model could potentially scale to 20 million people across the Middle East and North Africa (MENA). We believe this model can redefine how support is delivered, not only in Jordan, but globally.

If you’re interested in partnering, adapting this approach to your context, or supporting the next stage of research and scale, we’d love to connect.