Research library

Plain-language summaries of research, news, and policy on how AI is affecting addiction recovery, treatment, and research. Newest first.

1. Study · Chatbots and AI companions in recovery

   A recovery chatbot did no better than emailed education in a randomized trial

   Source date April 24, 2026 · Randomized trial, registered in advance (NCT04925570) · 202 analyzed of 258 randomized (107 chatbot, 95 email)

   In a randomized trial (202 adults analyzed), people using the Woebot substance use chatbot cut back about as much as people sent education by email.

2. Policy · Privacy, policy, and law

   Illinois makes it illegal for AI to act as your therapist

   Source date August 1, 2025

   An Illinois law bars AI from providing therapy and limits licensed clinicians to using AI for paperwork and support tasks. Fines reach $10,000 per violation.

3. Study · AI in the clinic

   A hospital AI screener for opioid use disorder matched usual care, and consulted patients were readmitted less often afterward

   Source date April 3, 2025 · Before-and-after comparison at one hospital, designed as a non-inferiority test. Not randomized. Registered as NCT05745480. · 51,760 adult hospitalizations

   An AI tool reading hospital notes matched usual care on addiction consults; consulted patients had fewer 30-day readmissions afterward. Not a randomized trial.

4. Study · Chatbots and AI companions in recovery

   ChatGPT-4 gave sound answers to common alcohol questions, but rarely pointed people to help

   Source date August 14, 2024 · Content evaluation. Two coders rated every answer against National Institute on Alcohol Abuse and Alcoholism (NIAAA) resources. Not pre-registered; the authors call the results exploratory. · 64 questions, each asked once in a new chat

   ChatGPT-4 answered 64 common questions about alcohol use disorder. 59 answers were fully evidence-based, but only 8 pointed people to outside help.

5. Study · Chatbots and AI companions in recovery

   Chatbots answered real recovery questions well overall, and some answers were dangerous

   Source date June 26, 2024 · Evaluation of chatbot answers. Clinicians rated them without being told the answers came from AI. · 75 questions, 150 answers, rated by 7 clinicians

   Clinicians rated ChatGPT and LLaMA-2 answers to recovery-forum questions as high quality, but some gave wrong helplines or endorsed home detox.

6. Policy · Privacy, policy, and law

   The federal privacy rule for addiction treatment records now works more like HIPAA

   Source date February 16, 2024

   HHS revised 42 CFR Part 2, the privacy rule for SUD treatment records, to align it with HIPAA. Programs had to comply with most provisions by February 16, 2026.

7. News · Digital therapeutics and apps

   The first FDA-authorized app for substance use disorder, and its maker's bankruptcy

   Source date April 7, 2023 · Multi-site, unblinded, randomized trial run through the National Institute on Drug Abuse Clinical Trials Network (CTN0044), 12 weeks · 507 in the trial. The FDA's intended-use figures come from the 399 who did not name opioids as their main substance.

   In 2017 the FDA let Pear market reSET, the first mobile medical app to help treat substance use disorders. In April 2023, Pear filed for Chapter 11.

8. Study · AI in addiction research

   An AI search for cocaine addiction medicines led experts to ketamine, a drug already being studied

   Source date February 23, 2023 · AI ranking of 1,430 FDA-approved drugs, review by a 7-member expert panel, then a retrospective health-record comparison with propensity score matching. Not pre-registered; the authors call it exploratory. · 7,742 patients given anesthesia (3,871 ketamine, 3,871 other anesthetics) and 7,910 patients with depression (3,955 ketamine, 3,955 antidepressants or midazolam), after matching

   Experts chose ketamine from an AI shortlist for cocaine addiction. Records tied it to more remission diagnoses, but just 1% to 3% of ketamine patients had one.

9. Study · Predicting relapse and overdose

   An overdose risk score held up in a second state, but most people it flagged did not overdose

   Source date June 1, 2022 · Prediction model built and tested on past insurance claims (a prognostic modelling study). No one was treated differently because of the score. · Built on 639,693 Pennsylvania beneficiaries (2013-16). Tested on 318,585 Pennsylvania beneficiaries (2017-18) and 391,959 Arizona beneficiaries (2015-17).

   A Medicaid claims model ranked opioid overdose risk well in Pennsylvania and Arizona. In its top Pennsylvania risk group, about 4 in 100 had an overdose.