The Loneliness Loop: A Therapist Guide to AI Companion and Chatbot Dependence
Behavioural Addiction · Practice-informed · Practice-informed
An integrative third-wave protocol for clients whose main source of emotional connection has become an AI chatbot or companion app. It formulates the pattern with the components model of behavioural addiction (salience, mood modification, tolerance, withdrawal, conflict and relapse), then sequences compassion-focused work before ACT values work, followed by DBT interpersonal effectiveness and regulation skills and a schema-informed relapse prevention session. A 2026 cross-sectional study of young adults found a strong link between loneliness and chatbot dependence, partly explained by low perceived social support, which gives clinicians a working formulation to test with each client. There are no published trials of treatment for this presentation yet, so the overall evidence rating is practice-based, the lowest tier present, and each technique is rated honestly for this population.
Underlying mechanism: Integrative Protocol for AI Companion Dependence: Components Model Formulation, CFT, ACT and DBT Skills
What it helps with
- Giving clinicians a shared, non-shaming language for a new presentation, using the six components of behavioural addiction as a self-assessment rather than a label.
- Separating heavy but adaptive AI companion use, such as rehearsing social scripts or short-term grief support, from use that is crowding out human relationships or a client's own values.
- Softening shame and self-criticism about the pattern first, so that values work lands more openly and secrecy is less likely to keep the behaviour going.
- Replacing AI-seeking with practisable alternatives for both emotion regulation and connection, so behaviour change does not simply leave a gap.
- Planning for lapses and widening human contact at a pace the client can sustain, with a booster session built in.
How clinicians use it
- Assess function before frequency, and build the case for change from the client's own values and functioning rather than from hours of use.
- Run the five sessions in order: formulation, the CFT session, the ACT session, DBT skills, then relapse prevention and reconnection, with a booster six to eight weeks after the block.
- Keep the CFT session before the ACT session by design. Softening self-criticism and threat-system activation first tends to leave more room for open values work.
- Teach nose-low-slow breathing first and offer resonance breathing, four seconds in and six out, only later as an optional second technique.
- Complete your own risk assessment and, for clients under 18, assess exposure to romantic or sexualised companion content as a safeguarding matter in its own right.
Clinical Implementation Guide
- Assess function before frequency, and screen for differential and comorbid presentations before starting.
- Keep the compassion-focused session (2) before the ACT values session (3) so shame is softened first.
- Schedule a booster session six to eight weeks after the block ends, given how new this presentation is.
What the worksheet looks like
| Session | Focus | Worksheet(s) |
|---|---|---|
| 1 | Engagement and formulation: the components model, what the companion gives and costs, and a start on the urge diary | W1, W5 |
| 2 | Compassion-focused work: the threat, drive and soothe systems, and soothing rhythm breathing | W2 |
| 3 | ACT values and workability: away moves, toward moves and defusion | W3 |
| 4 | DBT skills: interpersonal effectiveness, alternatives to the app and urge surfing | W4, W5 |
| 5 | Relapse prevention: the underlying need, a graded reconnection ladder and a lapse plan | W6 |
Clinical cautions
AI chatbots are not a reliable source of risk assessment or crisis support, so a client's companion app must never stand in for your own risk assessment. Screen for autism spectrum presentations, social anxiety, depression, avoidant or disorganised attachment, complicated grief and psychosis, and treat unusually fixed beliefs about the AI's sentience or feelings as a separate clinical concern. Distinguish avoidance that AI use has enabled from structural isolation it is compensating for. This tool does not diagnose any condition and is not a substitute for individual face-to-face therapy or for emergency crisis support. Professional judgement and supervision apply throughout.
References
- Griffiths, M. D. (2005). A "components" model of addiction within a biopsychosocial framework. Journal of Substance Use, 10(4), 191-197. https://doi.org/10.1080/14659890500114359
- Uğuz, Ö., Doğan, S., & Güngör, D. (2026). Loneliness and AI chatbot dependence in young adults: The mediating role of perceived social support. BMC Psychology, 14, 1184. https://doi.org/10.1186/s40359-026-04916-5
- Packer, K., Kingston, J. L., & Tibber, M. S. (2026). Acceptance and Commitment Therapy (ACT) interventions for digital technology-related issues: A scoping review. Journal of Contextual Behavioral Science, 41, 101003. https://doi.org/10.1016/j.jcbs.2026.101003
- Vidal, J., & Soldevilla, J. M. (2023). Effect of compassion-focused therapy on self-criticism and self-soothing: A meta-analysis. British Journal of Clinical Psychology, 62(1), 70-81. https://doi.org/10.1111/bjc.12394
- Lehrer, P. M., & Gevirtz, R. (2014). Heart rate variability biofeedback: How and why does it work? Frontiers in Psychology, 5, 756. https://doi.org/10.3389/fpsyg.2014.00756
Part of
The Loneliness Loop
Related worksheets
Next steps
- Browse the full clinical resource library — every worksheet, free and printable.
- Hear about new free resources. Join the clinician email list for occasional notes when there is something useful to share. The sign-up form is at the foot of this page once it has finished loading. This worksheet is free without signing up; unsubscribe anytime.
- Professional and Studio workflows — saved Client Packs, secure delivery and programme publishing when you need more than individual free resources.
