Helping people access drug and alcohol information safely

When people have questions about alcohol and other drugs, finding reliable answers isn't always straightforward. The Alcohol and Drug Foundation (ADF) identified 3 persistent barriers stopping people from getting the information they need:

• Information overload: the volume of alcohol and other drug information available online can be overwhelming, and quality varies widely. Many people don't know where to look or how to judge whether information is credible.
• Stigma: many people feel uncomfortable raising questions about drug use with a health professional or another person.
• Health literacy: health information can be difficult to read and understand, which can stop people from seeking help.

The ADF needed a way to provide fast, trustworthy, easy-to-understand information, without judgment. 

The ADF built dib, the drug info bot. It’s a friendly, AI-powered information chatbot available through the ADF website. Anyone can message dib their questions about alcohol and other drugs in a chat-based format. Conversations are always free, anonymous and secure.

Critically, dib draws its answers exclusively from the ADF's own evidence-based website content, so the information it provides is reliable and consistent with clinical guidance. ADF’s Ethical Framework for Working with AI and the data practices used for dib also aligned with the Australian Government’s Voluntary AI Safety Standards (now Guidance for AI Adoption).  

Creating dib took a careful, staged approach:

• An expert advisory group was established, bringing together AI specialists, academic and clinical leaders, and people with lived experience of alcohol and drug use from across Australia. This group shaped the ethical framework, governance model, evaluation methodology and risk mitigation strategies.
• Dib's responses were extensively tested against human subject matter experts at the ADF to verify accuracy and refine tone.
• A clinical advisory group reviewed outputs from a clinical standpoint and established safety guardrails.
• The ADF consulted with a peak harm reduction organisation and the broader treatment sector.
• Thorough user testing was conducted with members of the Australian public before launch.

After launch, the ADF put strong safeguards in place:

• A unique protection system detects crisis situations and automatically provides links to emergency services.
• Robust guardrails protect against misuse.
• A ‘human-in-the-loop' alert system means all alerts are reviewed by humans, alongside random conversation samples. This creates an ongoing evaluation dataset for that is used for continuous improvement. 

Since launching just over a year ago, dib has had more than 47,500 free and anonymous conversations, helping tens of thousands of Australians quickly access reliable information on alcohol and drugs.

People aged 18–24 are the most frequent users of dib, accounting for more than half (53%) of all conversations.

Alcohol is the most common topic. People ask about alcohol two to three times more per month than cannabis, which is the second most common topic.

Many people use dib to find out how they can reduce their drinking or access support. This suggests that dib is reaching people at a meaningful point in their journey seeking help.

By using AI to deliver the ADF's trusted content at scale, dib is reaching people who may never have felt comfortable asking another person for help. 

• Start with the problem, not the technology. The ADF's starting point was a clear understanding of why people weren't getting help: information overload, stigma and health literacy barriers. AI was the solution to a defined human need.
• Safety and governance aren't optional extras. Building dib involved an expert advisory group, a clinical advisory group, sector consultation and user testing. This investment in the process is what makes the tool trustworthy.
• Ground AI in authoritative content. By limiting dib's responses to the ADF's own evidence-based content, the organisation ensured consistent, reliable answers – and minimised the risk of the AI generating inaccurate or harmful information.
• Ensure ongoing human oversight. Automated alerts, random conversation sampling and regular human review mean the ADF can identify issues early and keep improving the tool over time.
• Reach the people who need it most. More than half of dib's users are aged 18-24 – a cohort that can be hard to engage through traditional health channels. AI-powered tools, done well, can open new pathways to support.