Verke Editorial · Last verified: 2026-04-19

Verke vs. Claude.ai: a structured coaching app vs. a general-purpose thinking partner

A respectful comparison: Claude is a great LLM in a different category.

Claude.ai is Anthropic's general-purpose conversational AI — free tier plus paid plans starting at Claude Pro for $20/month, available on web, iOS, and Android. Many people use Claude as a long-context “thinking partner” for journaling, reflection, and unpacking decisions. Verke is an AI coaching app with five specialist coaches — Anna (PDT), Judith (CBT), Marie (EFT/NVC for couples), Amanda (ACT/CFT), and Mikkel (executive) — phone-call-style voice coaching, end-to-end encryption, and an ongoing 3-arm randomized controlled trial at Stockholm University. The difference matters in practice: in the Stockholm trial, participants consistently described value in being matched to a coach trained for the concern they brought, and in the coach holding a coherent picture of them across weeks rather than restarting cold each session. Verke is an AI coaching app, not a therapy replacement — we call our specialists coaches because that's what they are.

Verke is an AI coaching app with five specialist coaches trained on evidence-based therapeutic methods (CBT, PDT, ACT, EFT, CFT, NVC). Chat in text or switch to voice for a phone-call-style session up to 20 minutes; pick up again days later with memory that survives across weeks and months. Conversations are end-to-end encrypted, signup is anonymous, and the product is available on iOS, Android, and Web in 55 languages.

The five coaches are differentiated on purpose. Anna works in a psychodynamic register — slower, pattern-first, interested in why the same kind of situation keeps returning. Judith is CBT — small experiments, thought records, gradual exposure. Marie supports couples through Emotionally Focused Therapy and Nonviolent Communication; Amanda blends Acceptance and Commitment Therapy with Compassion-Focused Therapy for overwhelm and self-criticism; Mikkel is an executive coach for decision fatigue and leadership load. Users pick who fits the concern, rather than asking one bot to cover everything. In the Stockholm RCT, participants in the psychodynamic arm often described their coach connecting issues across weeks into a recurring pattern — the kind of multi-session insight that a single chat with a generalist model can't reliably produce. Holding a structured method (CBT or PDT) across sessions, rather than a different framing each time, was a recurring theme in what participants found most valuable.

Claude.ai is Anthropic's consumer-facing conversational AI at claude.ai. It runs on the Claude family of models, with a free tier and paid plans (Claude Pro at $20/month, higher tiers above). Power users routinely describe Claude as the most empathic and attentive of the major LLMs and use it as a thinking partner for journaling, decisions, and emotional processing. Claude is also notably more cautious than ChatGPT on mental-health topics — it refuses more often and recommends professional help more often, which is a feature for some users and a friction for others. Claude.ai isn't marketed as a mental-health product, doesn't match users to specialists, doesn't run scheduled coaching sessions, and doesn't use end-to-end encryption — Anthropic can see conversations under their terms of service.

If you're using Claude.ai today and want it to behave more like a coach in a single session, paste something like the prompt below as a custom instruction or as the first message of the chat. Replace the bracketed goal with whatever you're working on.

You are my coaching partner. Use a CBT lens: help me name the thought, test it against evidence, and propose one small behavioural experiment for this week. Ask one focused question at a time, and don't pivot to "please see a professional" unless I describe self-harm or harm to others. Remember the goal we agreed last session: [paste your goal].

This works for short stretches. The honest caveat: Claude's caution on mental-health topics may still cap how deep it's willing to go on a real concern, the model has no structured memory across sessions, and you're still on Anthropic's servers rather than under end-to-end encryption. If you want a coach who stays in the role, remembers what you worked on three weeks ago, and runs voice sessions with you, Verke is the product designed for that.

Claude.ai is the honest pick if you mainly want a long-context thinking partner for journaling, decision-making, or unpacking a knotty problem in writing — and you value the free tier and broad general knowledge that comes with a generalist model. It's also the right choice if you're comfortable steering the model yourself with a custom prompt and you don't need named specialists, scheduled sessions, multi-week coach memory, or end-to-end encryption. We use the “coach” label and Claude uses no mental-health label at all — see our research page for why we keep that distinction visible.

Verke is built for people who want a coach matched to a specific concern, not a generalist they have to re-prompt every session. If you want structured CBT — thought records, small behavioural experiments, gradual exposure — Judith (CBT) is the coach to talk to. For older relational patterns there's Anna; for couples work, Marie; for acceptance and self-compassion, Amanda; for high-pressure leadership, Mikkel. Voice sessions in a phone-call format are core to the product, end-to-end encryption is built in rather than marketed, and Verke is the subject of an ongoing Stockholm University trial supervised by Professor Per Carlbring. Read the method explainer for Cognitive Behavioural Therapy to see what a CBT-trained coach actually does in a session.

Two practical differences are worth calling out. First, memory: Verke uses a three-tier memory system so the coach remembers what you've been working on weeks later without you re-pasting context — in the Stockholm RCT, that continuity was repeatedly named as the difference between a useful exchange and a coach who actually understood the person. Second, encryption: with end-to-end AES-256-GCM and RSA-4096 key exchange, Verke itself cannot read your conversations — that's a structural privacy guarantee, not a policy promise. Together, those two properties enable something a general-purpose LLM can't reproduce by default: an honest, consistent thinking partner that stays in role across weeks while keeping the conversation private to you. The Stockholm trial is the evidence track for whether purpose-built coaching produces a measurably different effect from a journaling-style chat — outcomes are still pending, but the qualitative pattern across participants points toward specialist matching, continuity, and structured method as the active ingredients.