{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,19]],"date-time":"2026-09-19T04:15:44Z","timestamp":1789791344569,"version":"4.0.1"},"reference-count":93,"publisher":"JMIR Publications Inc.","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["JMIR Ment Health"],"abstract":"<jats:p>The rapid evolution of large language models has accelerated the development of agentic artificial intelligence (AI) systems capable of pursuing autonomous goals, creating an urgent need for structural frameworks in psychiatry and psychotherapy. While existing classifications often draw parallels to autonomous driving, this paper argues that the mental health domain requires a distinct, domain-specific theoretical foundation, as the 2 domains differ fundamentally in their semantic, ideographic, and epistemological demands. Furthermore, they differ in their end goals, for which we introduce terms such as agentic guidance capability. To guide clinicians and researchers through these developments, we propose a 5-stage taxonomy for language-based AI systems that differentiates technical functionality from clinical effectiveness. The taxonomy progresses from level 1 (knowledge level), in which systems perform static benchmark tasks, to level 2 (elementary level), characterized by dynamic engagement in specific therapeutic microskills. At level 3 (integration level), systems achieve consistency across and within modules, as well as basic case-level conceptualization suitable for blended therapy under human oversight. Level 4 (saturation level) describes therapist-in-the-loop systems capable of autonomous functioning with minimal supervision, whereas level 5 (mastery level) represents AI systems that are technically capable of performing autonomous therapy. By distinguishing technical functionality from clinical effectiveness, we conclude that level 4 or level 5 performance does not automatically translate into full treatment effectiveness, even if high treatment fidelity can be achieved. We conclude by emphasizing the need to shift benchmarking from static knowledge tests to dynamic evaluations of therapeutic capabilities in order to safely navigate the transition toward autonomous care.<\/jats:p>","DOI":"10.2196\/91746","type":"journal-article","created":{"date-parts":[[2026,4,18]],"date-time":"2026-04-18T19:45:07Z","timestamp":1776541507000},"page":"e91746","source":"Crossref","is-referenced-by-count":2,"title":["AI Agents Are Coming: 5-Stage Taxonomy of Language-Based AI Systems for Psychiatry, Psychotherapy, and Counseling"],"prefix":"10.2196","volume":"13","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-6447-9132","authenticated-orcid":false,"given":"Raphael","family":"Schuster","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4907-3505","authenticated-orcid":false,"given":"Constantin Yves","family":"Plessen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2172-8813","authenticated-orcid":false,"given":"Per","family":"Carlbring","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4516-1783","authenticated-orcid":false,"given":"Andreas","family":"Walther","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1010","published-online":{"date-parts":[[2026,7,13]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1002\/wps.21352"},{"key":"ref2","unstructured":"HuKChatGPT sets record for fastest-growing user base - analyst noteReuters2023-02-01https:\/\/www.reuters.com\/technology\/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01\/"},{"key":"ref3","author":"Taleb, NN","year":"2010","journal-title":"The Black Swan: The Impact of the Highly Improbable. 2nd Edition"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1093\/mind\/LIX.236.433"},{"key":"ref5","author":"Hawking, SW","year":"1988","journal-title":"A Brief History of Time: From the Big Bang to Black Holes"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1007\/s12160-016-9830-8"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.3389\/fdgth.2025.1460167"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1002\/wps.21342"},{"key":"ref9","unstructured":"RadfordAWuJChildRLuanDAmodeiDSutskeverILanguage models are unsupervised multitask learnersOpenAI Technical Report20192026-06-27https:\/\/cdn.openai.com\/better-language-models\/language_models_are_unsupervised_multitask_learners.pdf"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.32872\/cpe.12083"},{"key":"ref11","unstructured":"GodagerGObrizanMPompeoMCan AI help Ukraine's mental health crisis? New evidence on digital support during warVoxUkraine2025-12-04https:\/\/voxukraine.org\/en\/can-ai-help-ukraines-mental-health-crisis-new-evidence-on-digital-support-during-war"},{"key":"ref12","year":"2025","journal-title":"WHO's Response to Health Emergencies in Ukraine: Annual Report 2024"},{"key":"ref13","unstructured":"McCarthyJMinskyMRochesterNShannonCA proposal for the Dartmouth summer research project on artificial intelligenceACM Digital Library19552026-06-27https:\/\/dl.acm.org\/doi\/10.1609\/aimag.v27i4.1904"},{"key":"ref14","unstructured":"McCarthyJPrograms with common sense1959Proceedings of the Teddington Conference on the Mechanization of Thought ProcessesNovember 24-27, 1958Teddington, England7591"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1017\/s0269888900008122"},{"key":"ref16","author":"Sutton, RS","year":"1998","journal-title":"Reinforcement Learning: An Introduction"},{"key":"ref17","author":"Huhns, MN","year":"1989","journal-title":"Distributed Artificial Intelligence"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1145\/212094.212121"},{"key":"ref19","author":"Ferber, J","year":"1999","journal-title":"Multi-Agent Systems: An Introduction to Distributed Artificial Intelligence"},{"key":"ref20","doi-asserted-by":"crossref","DOI":"10.7551\/mitpress\/5236.001.0001","author":"Rumelhart, DE","year":"1986","journal-title":"Parallel Distributed Processing: Explorations in the Microstructure of Cognition. Vol 1: Foundations"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1037\/0033-295x.88.2.135"},{"key":"ref22","unstructured":"TambeMTeamwork in real-world, dynamic environments1996Proceedings of the Second International Conference on Multi-Agent Systems (ICMAS-96)December 9-13, 1996Kyoto, Japan"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1145\/544862.544946"},{"key":"ref24","unstructured":"AndrewNgAgentic design patterns part 1: four AI agent strategies that improve GPT-4 and GPT-3.5 performanceDeepLearning.AI2026-07-11https:\/\/www.deeplearning.ai\/the-batch\/how-agents-can-improve-llm-performance\/"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/TCCN.2026.3658762"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1177\/0278364920987859"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/tits.2021.3054625"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1038\/s44184-025-00159-1"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2410.06845"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2408.15787"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2503.16521"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.711"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.2196\/68056"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/3706598.3714014"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1016\/j.neubiorev.2016.06.022"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1016\/j.neubiorev.2025.106296"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1001\/jamapsychiatry.2017.0044"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1002\/wps.70002"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1001\/jamapsychiatry.2020.4364"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1001\/jama.2025.23965"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1056\/aioa2400802"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1016\/j.jad.2024.04.057"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1097\/YCO.0000000000001048"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1038\/s44184-024-00056-z"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.3390\/wevj15060262"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1126\/sciadv.aaw2594"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/taffc.2015.2457417"},{"key":"ref48","unstructured":"GourlayDWhitelawBComparing datasets for AI tuning and inference: large language models vs autonomous driving systemsQumulo2026-03-27https:\/\/qumulo.com\/blog\/ceo\/comparing-datasets-for-ai-tuning-and-inference-large-language-models-vs-autonomous-driving-systems\/"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1257\/pandp.20181019"},{"key":"ref50","author":"Susskind, D","year":"2020","journal-title":"A World Without Work: Technology, Automation, and How We Should Respond"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.2196\/18100"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1016\/j.cpr.2022.102179"},{"key":"ref53","unstructured":"Terence Tao at IMO 2024: AI and mathematicsYouTube2026-03-27https:\/\/www.youtube.com\/watch?v=e049IoFBnLA"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1911.01547"},{"key":"ref55","unstructured":"ARC Challenge2026-06-17https:\/\/arcprize.org\/"},{"key":"ref56","author":"Wang, S","journal-title":"ArXiv. Preprint posted online on August 11, 2025"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2311.12022"},{"key":"ref58","unstructured":"GPQA benchmark leaderboard: testing LLMs on graduate-level science questionsBRACAI Consulting20242026-03-27https:\/\/www.bracai.eu\/post\/gpqa-benchmark-leaderboard"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2508.15361"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2505.08775"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.2196\/58195"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2509.04183"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2506.08584"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.naacl-long.196"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1145\/3711896.3737393"},{"key":"ref66","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.830"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2502.15418"},{"key":"ref68","unstructured":"HuangFChbeirSWangSTanSLouieRDanielMAdeliETherapyGym: evaluating and aligning clinical fidelity and safety in therapy chatbotsOpenReview20252025-08-19https:\/\/openreview.net\/forum?id=nwANzdlMKI"},{"key":"ref69","doi-asserted-by":"publisher","DOI":"10.1038\/s44277-025-00049-6"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.1109\/fmlds67896.2025.00067"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2026.eacl-long.180"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1016\/j.invent.2025.100841"},{"key":"ref73","unstructured":"Humanity's last examCenter for AI Safety20252025-12-29https:\/\/agi.safe.ai\/"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2501.14249"},{"key":"ref75","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2501.09426"},{"key":"ref76","doi-asserted-by":"publisher","DOI":"10.1016\/j.invent.2025.100882"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.1145\/3341981.3344241"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.1016\/j.aiopen.2024.01.003"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2509.00559"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2403.11901"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2601.06152"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2511.10652"},{"key":"ref83","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2506.12606"},{"key":"ref84","author":"Chen, A","journal-title":"ArXiv. Preprint posted online on February 12, 2026"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2602.20648"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2602.21638"},{"key":"ref87","unstructured":"Genie 2: a large-scale foundation world modelGoogle DeepMind20242026-04-29https:\/\/deepmind.google\/discover\/blog\/genie-2-a-large-scale-foundation-world-model\/"},{"key":"ref88","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-025-14923-y"},{"key":"ref89","doi-asserted-by":"publisher","DOI":"10.1016\/j.brat.2025.104857"},{"key":"ref90","doi-asserted-by":"publisher","DOI":"10.1080\/10503307.2025.2539405"},{"key":"ref91","doi-asserted-by":"publisher","DOI":"10.1016\/S0140-6736(25)01237-1"},{"key":"ref92","doi-asserted-by":"publisher","DOI":"10.2196\/79658"},{"key":"ref93","doi-asserted-by":"publisher","DOI":"10.1016\/j.invent.2025.100824"}],"container-title":["JMIR Mental Health"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/mental.jmir.org\/2026\/1\/e91746","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,13]],"date-time":"2026-07-13T15:45:14Z","timestamp":1783957514000},"score":1,"resource":{"primary":{"URL":"https:\/\/mental.jmir.org\/2026\/1\/e91746"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,13]]},"references-count":93,"URL":"https:\/\/doi.org\/10.2196\/91746","relation":{"has-preprint":[{"id-type":"doi","id":"10.2196\/preprints.91746","asserted-by":"object"}]},"ISSN":["2368-7959"],"issn-type":[{"value":"2368-7959","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,13]]}}}