Artificial intelligence is rapidly reshaping healthcare, with growing potential not just in clinical settings but also in supporting how people change their behaviour. Many of the most persistent health challenges are behavioural in nature and smoking is one of the most harmful examples. Despite sustained public health efforts and access to effective treatments, tobacco use remains a leading cause of preventable disease and premature death worldwide, causing the deaths of more than 7 million people each year according to the World Health Organisation. In the UK alone, smoking still accounts for around 75,000 deaths annually and remains a major driver of health inequality.
The Real Barrier to Quitting
For most people who smoke, the barrier is not information but the moment when stress, habit or social cues override long term intentions. Relapse risk can often be heightened around predictable contexts and emotional states. This is where AI has a pivotal role to play, complementing the work of clinicians to provide support at the moment it is needed.
Almost all behaviour change is difficult, smoking particularly so. On average, it takes around 30 attempts for someone to quit successfully, and roughly 97 percent of unsupported quit attempts do not succeed. That might seem surprising when you consider that most people who smoke want to stop, and that not stopping carries a very real risk of serious illness and early death.
The explanation is not lack of willpower. In fact, framing quitting as a test of willpower may be part of the problem. The real barriers lie in basic human psychology and the way habits become embedded in our minds.
Why Smoking Is So Hard to Stop
Two dynamics make smoking unusually difficult to stop. The first is the sheer number of cravings that must be resisted.
For many people who smoke the habit has become woven into almost every aspect of daily life. A typical smoker will consume well over one hundred thousand cigarettes before they quit. Smoking becomes linked to everyday routines like the first coffee of the day, the end of a meal, a stressful moment at work, or a quiet moment alone.
Smoking accompanies happiness, boredom, anxiety, celebration, and just about every other emotion we experience. Over time these associations become programmed into the brain and can automatically trigger the urge to smoke.
Someone who has decided to quit may begin with strong motivation, but motivation fluctuates. Even the most determined person will experience moments when it dips. When a moment of low motivation coincides with a deeply ingrained trigger, the urge to smoke can feel overwhelming.
The problem being that a single cigarette almost always leads to a full relapse back to smoking. The challenge is not simply deciding to stop but sustaining that decision across hundreds of vulnerable moments. What people need in those moments is immediate support. Help that is available at any time of day or night, wherever they happen to be.
The Limits of Traditional Support
The second difficulty is that the kind of support that works varies enormously between individuals and even from one situation to the next. A technique that helps someone manage a craving one day may be far less effective the following day.
Experienced therapists understand this intuitively. They recognise different personality types and know which behavioural techniques are more likely to resonate. They pick up on subtle cues about what someone is experiencing and adjust their approach in real time.
This is highly skilled work, and it does not scale easily. There simply are not enough trained behavioural specialists to meet the size of the problem.
And that problem is vast. Diseases linked to behaviours such as smoking, alcohol use, poor diet and physical inactivity account for more than 70 percent of premature deaths worldwide. If we want to reduce that burden meaningfully, we need approaches that can reach far more people.
Where AI Can Make a Difference
This is where AI starts to show real value. It is already being used across health tech to deliver personalised behavioural support at scale, with growing evidence of its effectiveness.
For years, behavioural science has been aiming for something that is easy to describe but difficult to deliver: the right technique, in the right format, for the right person, at the right moment. Until recently, that level of responsiveness has largely been limited to one-to-one support.
That is starting to change. Advances in AI mean personalised support at scale can now be delivered in practice.
A Practical Example of AI in Action
Systems can combine established evidence on smoking cessation with personal health data, shared with explicit user consent, from devices such as smartwatches. They can learn from people with similar behavioural profiles, smoking history, smoking triggers, and previous quit attempts. Used responsibly, this makes it possible to identify higher risk moments and offer timely, evidence-based support.
Take someone on day four of a quit attempt. It is a point where people often feel encouraged by their progress, but where relapse risk is still high and vigilance matters. Their smartwatch might show elevated stress levels, a well established predictor of relapse.
Their calendar shows they are about to leave for a meeting that will take them past several shops selling cigarettes. Based on this combination of context and past behaviour, the system flags a vulnerable moment. It prompts them to use a nicotine patch and carry nicotine gum to reduce the immediate pressure on willpower, and reminds them to use an if-then technique that has worked for them before.
The support is straightforward, but timed to when it is most needed. The building blocks for this approach already exist. The question now is how they are used, with clear safeguards around consent, privacy and clinical oversight.
Beyond Smoking
And this goes beyond smoking. Similar challenges exist in reducing alcohol use, improving diet, or becoming more physically active. If AI can support behaviour change in one area, there is every reason to expect similar approaches could work elsewhere.
Trust, Ethics and Adoption
Trust will be central to success. People need to understand how their data are used and feel confident they remain in control of it. Without that, even the most well-designed system will struggle to engage people, let alone change behaviour.
That makes the role of ethics and transparency critical. The organisations that take this seriously, and build it into their systems are likely to be the ones that succeed.
The Wider Behavioural Landscape
It is also worth remembering why behavioural health challenges became widespread in the first place. Many industries are highly effective at exploiting the same psychological mechanisms that make habits hard to break. Tobacco, alcohol, gambling, fast food, social media and gaming companies have invested heavily in understanding how to trigger desire, reward and compulsive engagement.
Advances in neuroscience, behavioural science and data analytics have allowed corporations to refine these techniques with increasing precision. Products and experiences are often designed not simply to meet existing demand but to reinforce and deepen it.
Resisting this requires learning a skill that humans have not historically needed to rely upon to such a degree. Our evolutionary systems are tuned to respond quickly to rewards and impulses because, for most of human history, those instincts were advantageous. In the modern environment they can work against us.
The Role of AI in Behavioural Health
AI can help millions of people find the right techniques to build their self-regulation muscles. AI is already working alongside human support, helping to personalise interventions, and address health inequalities in ways that would not otherwise be possible.
While AI cannot replicate human empathy, it can be trained to deliver consistent, supportive and stigma-aware communication. For some people, this can feel easier to engage with, particularly where fear of judgement is a barrier.
By supporting routine aspects of care, AI can help extend reach, allowing human specialists to dedicate more time to cases that require deeper clinical expertise and personalised intervention.
A More Equitable Future?
Often the people most affected by smoking and other harmful behaviours are also those facing the greatest social and economic challenges. Directing human expertise toward these individuals, while enabling AI to support others at scale, could make support systems more equitable, effective, and sustainable.
Whether this future materialises is not guaranteed. Technology alone rarely solves complex social problems. But the possibility now exists in a way that it has not before.
For those of us who have spent decades working on behavioural health, that possibility is difficult to ignore. The opportunity to help millions of people overcome habits that shorten their lives and damage their wellbeing is too important to dismiss.
It won’t be easy, but the potential to transform lives makes it an opportunity we must embrace.
