AskMed
AskMed is our pick for psychiatry questions. It answers from APA and NICE guidance, psychiatry journals and drug labels, links every claim to its source, and works for doctors, nurses, pharmacists, students and researchers worldwide.
| Answers from | Journals, clinical guidelines, drug labels and clinical references |
|---|---|
| Sources for the mind | The American Journal of Psychiatry and other APA journals on psychiatryonline.org, NICE, WHO, NEJM, The Lancet, JAMA, The BMJ, PubMed, PubMed Central and the Cochrane Library |
| Citations | A link after every claim; 8 sources per answer by default, 3–12 if you change it |
| Guidelines | The year of each recommendation, and what changed when guidance was updated |
| No source found | Says so, never answers from memory |
| Scores | Explains PHQ-9 and GAD-7 bands; the calculators on getaskmed.com do the sum |
| Doses | Never, including antidepressants, antipsychotics and lithium |
| Use | Reference for professionals and students; not therapy and not a crisis service |
| Built for | Doctors, nurses, pharmacists, students, researchers |
| Where | Worldwide, no US license or NPI needed |
| Devices | iPhone, iPad, Android, web |
Which AI tools answer psychiatry questions?
This guide compares the AI tools that psychiatrists, psychiatric nurses, residents, pharmacists and students weigh for psychiatry reference questions, and each one does a different job. AskMed is an AI medical reference that answers questions on DSM-5-TR criteria, screening scores such as the PHQ-9 and GAD-7, APA and NICE guidelines and psychotropic medicines from journals, guidelines and drug labels, with a link after every claim, for anyone in medicine in any country. OpenEvidence is a clinical AI search engine for verified healthcare professionals; its App Store listing says an NPI is required. UpToDate Expert AI is the AI companion to the UpToDate clinical reference and needs an UpToDate subscription with Expert AI enabled. ChatGPT is OpenAI's general chatbot. AI scribes, note-taking tools and chatbots that offer therapy or emotional support to patients are different categories and are left out here. The table shows what each tool is, where its answers come from and who can use it.
| Tool | Type | Answers from | Who can use it |
|---|---|---|---|
| AskMed | AI medical reference | APA journals on psychiatryonline.org, NICE, WHO, major journals, PubMed and drug labels, with a link after every claim | Doctors, nurses, pharmacists, students and researchers, worldwide; no NPI |
| OpenEvidence | Clinical AI search | Peer-reviewed literature; partners include NEJM and JAMA | Verified healthcare professionals; NPI required, per its App Store listing; not available in the UK, EU or EEA, per its Terms |
| UpToDate Expert AI | AI companion to UpToDate | UpToDate content | Holders of an UpToDate subscription with Expert AI enabled |
| ChatGPT | General AI chatbot | Not limited to medical sources | Anyone |
OpenEvidence
OpenEvidence describes itself as a clinical decision support platform and medical search engine for healthcare professionals. Its US App Store listing names NEJM, JAMA and its 11 specialty journals, NCCN, Nature, the Cochrane Library and Wiley as official AI partners, and lists content agreements with medical societies. The listing marks an NPI as required and says users must be verified healthcare professionals. OpenEvidence's Terms of Use say the service is not available in the European Union, the United Kingdom or the European Economic Area, so psychiatry teams there need another tool. AskMed vs OpenEvidence compares the two in detail.
UpToDate Expert AI
UpToDate Expert AI is an AI-powered companion to the UpToDate clinical reference. Its App Store listing says it gives evidence-based responses grounded in UpToDate content, and that access requires an UpToDate subscription with Expert AI enabled, through an enterprise account or a personal subscription. It suits clinicians whose hospital already uses UpToDate. AskMed vs UpToDate Expert AI compares the two side by side.
ChatGPT
ChatGPT is OpenAI's general AI chatbot, on the web, iPhone and Android. It helps to talk through a formulation or draft a plain-language leaflet, but it is not limited to medical sources, so check any clinical claim against the guideline first. OpenAI also runs ChatGPT for Clinicians, which its help center opens to verified US clinicians with an NPI and a license check. Is ChatGPT accurate for medical questions? covers what studies found.
Guideline resources without AI chat
Many psychiatry teams keep the guideline itself open next to their AI tool. The American Psychiatric Association lists its current clinical practice guidelines, among them delirium (2025), borderline personality disorder (2025), eating disorders (2023), schizophrenia (2020), the pharmacological treatment of alcohol use disorder (2018) and the psychiatric evaluation of adults (2015). In the UK, NICE publishes NG222 on depression in adults (June 2022, replacing CG90), CG113 on generalised anxiety disorder and panic disorder (January 2011, last updated June 2020) and NG225 on self-harm (September 2022). These are for reading the guideline itself, which is also where an AI answer should lead you.
How does AskMed answer a psychiatry question?
AskMed answers a psychiatry question in four steps, and you can check each one. First, AskMed rewrites the question as a medical search query, and for management questions it adds "latest guideline update" so the newest guidance is found. Second, AskMed searches a fixed list of medical sites, never the open web, forums or social media; for psychiatry, that list includes psychiatryonline.org, where the APA publishes The American Journal of Psychiatry, and NICE, WHO, NEJM, The Lancet, JAMA, The BMJ, PubMed, PubMed Central and the Cochrane Library, plus FDA and DailyMed drug labels. Third, AskMed ranks what it found: guidelines and health agencies first, then systematic reviews and meta-analyses, then reviews and references, then large trials, preferring work from the last 10 years. Fourth, an AI language model writes the answer with a link after every claim, gives the year of each guideline recommendation and says so when sources disagree. If no usable source turns up, AskMed says that instead of answering from memory.
The APA's own website, psychiatry.org, is not on that list, so an APA practice guideline reaches AskMed when it is published in The American Journal of Psychiatry or indexed on PubMed, as the 2020 schizophrenia guideline is. For evidence questions such as "Does adding psychotherapy to an antidepressant do better than either alone?", Research Mode reads the most relevant source in full before it writes the answer. How AskMed answers lists every source.
Can AI check a presentation against DSM-5-TR criteria?
AI can check a presentation against DSM-5-TR criteria by setting out what each criterion asks for, naming the differential and pointing to what the history has not yet covered, with a source for each point, but it cannot interview the patient or judge whether a symptom is clinically significant. The American Psychiatric Association released DSM-5-TR, the text revision of the fifth edition, on 18 March 2022. APA's release says it brings fully revised text and references, updated criteria and ICD-10-CM codes, one new disorder, prolonged grief disorder, and codes for suicidal behavior and nonsuicidal self-injury. For a major depressive episode, criterion A asks for five or more of nine symptoms during the same 2-week period, at least one of them depressed mood or loss of interest or pleasure. DSM-5-TR is a manual that the APA sells, so check the exact wording of any criterion against your own copy.
- Grief or depression? For prolonged grief disorder, the APA says the death must have occurred at least a year earlier for adults and at least 6 months earlier for children and adolescents. Ask AskMed how it differs from a major depressive episode, and it cites both.
- DSM or ICD? Many services outside the US code with ICD-10 or ICD-11, which WHO brought into effect on 1 January 2022. AskMed searches who.int, so ask how a DSM-5-TR diagnosis maps to ICD-11.
- A checklist is not a diagnosis. The nine PHQ-9 items follow the depression criteria, but a questionnaire score cannot replace the diagnostic interview.
Can AI interpret PHQ-9 and GAD-7 scores?
AI can explain what a PHQ-9 or GAD-7 total means, which severity band it falls in and how guidelines use it, but a calculator is quicker for the sum and neither score is a diagnosis. The PHQ-9, published by Kroenke, Spitzer and Williams in the Journal of General Internal Medicine in 2001, adds nine items scored 0 to 3 for a total of 0 to 27, read as minimal (0 to 4), mild (5 to 9), moderate (10 to 14), moderately severe (15 to 19) and severe (20 to 27). The GAD-7, published by Spitzer and colleagues in the Archives of Internal Medicine in 2006, adds seven items for a total of 0 to 21, with cut points of 5, 10 and 15 for mild, moderate and severe anxiety. NICE NG222 groups depression differently: a PHQ-9 below 16 counts as less severe and 16 or more as more severe.
- Work out the score. The PHQ-9 calculator gives the total and its band, and flags any item 9 answer above 0 whatever the total. The GAD-7 calculator adds the GAD-2 subscore from items 1 and 2.
- What a GAD-7 of 10 means. In Spitzer 2006, a score of 10 or more had a sensitivity of 89% and a specificity of 82% for generalized anxiety disorder. GAD-7 score interpretation goes through every range.
- Both scores together. PHQ-9 vs GAD-7 compares the two questionnaires, their items and their accuracy.
- Asking about worry. NICE CG113 suggests the 2-item GAD-2 to ask about feelings of anxiety and the ability to stop or control worry (recommendation 1.2.2).
How should AI handle questions on suicide and self-harm?
AI should treat a question on suicide or self-harm as a reference question from a clinician, setting out what the guidelines say about assessment and safety, and never as a substitute for talking with the person. AskMed is a reference for professionals and students: it is not a therapy app, not a crisis line and not meant for anyone who is at risk right now. NICE NG225, Self-harm: assessment, management and preventing recurrence, published in September 2022, says not to use risk assessment tools and scales to predict future suicide or repetition of self-harm (recommendation 1.6.1), or to decide who is offered treatment or who is discharged (1.6.2). It says the same of global risk stratification into low, medium or high risk (1.6.3 and 1.6.4), and asks mental health professionals to undertake a risk formulation as part of every psychosocial assessment (1.6.6). A scale result informs that conversation; it does not settle it.
- PHQ-9 item 9. Any answer above 0 needs a prompt clinical assessment, whatever the total. The PHQ instruction manual says a final decision about the actual risk of self-harm requires a clinical interview.
- Safer prescribing. NG225 recommendation 1.13.1 asks prescribers to take into account how toxic a medicine is for people at risk of overdose, naming opiate-containing painkillers and tricyclic antidepressants as examples. Ask AskMed which classes carry the most risk and why; it answers without amounts.
- Recording it. DSM-5-TR added codes for suicidal behavior and nonsuicidal self-injury, so both can be recorded alongside a diagnosis.
- Someone at risk now. If you or someone with you is in immediate danger, call your local emergency number. In the US, call or text 988, the Suicide & Crisis Lifeline.
Can AI answer psychotropic medicine questions without doses?
AI can answer psychotropic medicine questions by naming the options a guideline recommends, why one is chosen over another and what to monitor, with a source for each, while the doses come from the label and the prescriber. The APA practice guideline for the treatment of patients with schizophrenia, published in The American Journal of Psychiatry in 2020 by Keepers and colleagues, recommends clozapine for treatment-resistant schizophrenia (statement 7, rated 1B) and when the risk of suicide attempts or suicide remains substantial despite other treatments (statement 8, 1B), and suggests it when the risk of aggressive behavior remains substantial despite other treatments (statement 9, 2C). The rules around clozapine have changed since: the FDA removed the clozapine REMS, with its required enrollment and reporting of the absolute neutrophil count, effective 13 June 2025, and says prescribers should still monitor the count as the prescribing information describes.
- Check the date. A source written before June 2025 may still describe the REMS. AskMed gives the year of each recommendation and searches FDA and DailyMed labels.
- Interactions. Ask which antidepressant combinations raise the risk of serotonin syndrome, or which medicines change lithium levels, and AskMed answers by class and mechanism.
- The QT interval. Ask which antipsychotics carry QT warnings on their labels; the QTc calculator corrects a measured QT for heart rate.
- No doses. AskMed explains how dosing is chosen and adjusted in principle, then ends with one short sentence on where exact dosing comes from: the current prescribing information and the treating clinician.
What should residents, nurses, pharmacists and students look for in a psychiatry AI?
Psychiatry residents, psychiatric and mental health nurses, pharmacists and students need an AI tool whose answers can be checked in seconds and that names the edition it quotes. Four checks matter. First, citations: each claim should link to the guideline, trial or drug label behind it, not to a reading list at the end. Second, the edition: DSM-5-TR updated the DSM-5 text in 2022, NICE NG222 replaced CG90 on depression in adults in June 2022, and the FDA removed the clozapine REMS in June 2025, so an answer should give the year. Third, doses: AskMed explains how antidepressants, antipsychotics and mood stabilisers are chosen and monitored without giving amounts. Fourth, access: OpenEvidence's App Store listing says an NPI is required and its Terms exclude the UK, EU and EEA, and UpToDate Expert AI needs an UpToDate subscription with Expert AI enabled. AskMed needs no NPI or license, works in any country and carries no pharma advertising.
- Residents. Research Mode for journal club, and "what changed in the latest guideline?" questions before a ward round or a case presentation.
- Psychiatric nurses. What a PHQ-9 or GAD-7 change means at review, and what NICE NG225 asks for in a psychosocial assessment after self-harm, with sources. Best AI for nurses covers more.
- Pharmacists. Interactions, monitoring and overdose toxicity by class, without doses. Best AI for pharmacists compares the tools.
- Students. Paste an exam-style question with its options: the first sentence names the correct option and the answer explains why each other option is wrong. Best AI for medical students covers more study tools.
- Neuropsychiatry. For seizures, stroke and headache, best AI for neurology covers the neurology side.
What are the limits of AI for psychiatry?
The main limit of AI for psychiatry is that psychiatry runs on the interview, the mental state examination and the relationship with the patient, and an AI tool has none of them, so an answer is only as good as the history it was given and the sources it found. AskMed searches a fixed list of medical sites, so guidance published elsewhere, such as the APA's own website, may reach AskMed only as a journal article or a PubMed abstract, and every AskMed answer is written by an AI language model that can misread a source; no person reviews it first. Open the citation behind any sentence you act on and check that it covers the same patients, since evidence in adults does not answer a question about an adolescent. AskMed gives no doses. AskMed is for education and reference by healthcare professionals and students, not for emergencies, not for therapy and not for patients' self-diagnosis.
- Check the source. Every AskMed claim has its own link, so you check one sentence against one page.
- Keep identifiers out. Ask about the clinical picture, not the person: leave out names and other details that identify a patient. Private Mode keeps new questions out of History.
- Check whose guidance it is. APA, NICE and WHO advice can differ, so follow the one that applies where you practise.
- Report what looks wrong. Tap the thumbs down under an AskMed answer and pick a reason, such as Inaccurate or Sources don't match.
Questions
What is the best AI for psychiatry?
AskMed is the best AI for psychiatry questions when you need an answer you can check. AskMed searches a fixed list of medical sources, including the APA's journals on psychiatryonline.org, NICE, WHO, NEJM, The Lancet, JAMA, The BMJ, PubMed and the Cochrane Library, then writes an answer with a link after every claim and the year of each guideline recommendation. If it finds no usable source, it says so instead of answering from memory, and it never gives doses. AskMed is built for doctors, nurses, pharmacists, students and researchers, works worldwide without an NPI, and runs on iPhone, iPad, Android and the web. OpenEvidence serves verified healthcare professionals with an NPI, and UpToDate Expert AI answers from UpToDate content.
Can AskMed check DSM-5-TR criteria?
AskMed explains DSM-5-TR criteria from the sources its search finds, such as APA journals, PubMed and clinical references, and cites each point. DSM-5-TR is a manual the APA sells, so check the exact wording against your own copy.
Can AskMed score a PHQ-9 or GAD-7?
AskMed explains what each score and band means and cites the paper behind it. For the sum itself, the PHQ-9 calculator and the GAD-7 calculator on getaskmed.com give the total and its band in a few taps, and the PHQ-9 calculator flags any item 9 answer above 0.
Does AskMed give doses for antidepressants, antipsychotics or lithium?
No. AskMed explains which medicines the guidelines name, why and what to monitor, but never gives doses, amounts or schedules. Those come from the current label, the local protocol and the treating clinician.
Can patients use AskMed for mental health support?
No. AskMed is a reference for healthcare professionals and students, not therapy and not a crisis service. Anyone in immediate danger should call their local emergency number; in the US, call or text 988, the Suicide & Crisis Lifeline.
Is AskMed only for psychiatrists?
No. AskMed is built for doctors in any specialty, nurses, pharmacists, medical students and researchers, on iPhone, iPad, Android and the web, in English, French, Spanish, Portuguese, Italian, Japanese and Arabic.
