Upload all the evidence and information you want — conversations, incidents, the history as you see it. Independent AI models discuss it against the couples-research literature, debate each other, and send you their opinion: what they think is wrong, the possible causes, and the stepwise strategies most likely to succeed — numbered, in order. The consensus PDF lands in your inbox automatically. This is AI Council opinion on your situation — not couples therapy.
Stripe checkout. Pick your council size — $25 to $249.
On the next page: your question, your situation, and any PDFs. Held only in memory — never stored.
On AWS Bedrock under a signed HIPAA BAA — no outside model is ever contacted. Independent AIs open, critique each other, a chair writes the consensus.
The AI Opinion lands in your inbox automatically — usually within the hour. Then every copy of your data is gone.
Your documents and details are handled only in memory while the council debates, then discarded the moment your PDF is sent. No database, no folder, nothing to leak. The only record kept is the delivery email itself — fingerprinted (SHA-256) so the report's authenticity can always be proven and forgeries disproven.
Every AI model runs on AWS Bedrock under a signed Business Associate Agreement (BAA). No outside model is ever contacted; your case is sent only to those in-BAA models and is never stored.
Your material produces your report and nothing else — it is never used to train models, ours or anyone's.
100% automated end to end — never reviewed by any human being.
Card data goes to Stripe's hosted checkout and never touches our servers.
Everything you send travels over TLS-encrypted connections.
The purpose of the AI Council is to throw complex topics and discussions into debate between independent, integrated AI models and LLMs — positions, critique, consensus — and hand you the result, while protecting your security, confidentiality, and anonymity: every AI model runs on AWS Bedrock under a signed HIPAA Business Associate Agreement (BAA), no outside model is ever contacted, and nothing you send is ever stored — it is held in memory only while the council debates, then discarded. Every AI Council product is 100% automated, never reviewed by any human being, and never used for training, ours or anyone's.
The same council that runs our medical and legal checks — models scoring ~92% on medical licensing (MedQA) and 82–84% on legal reasoning (LegalBench) — applied to the couples-research literature (Gottman, attachment, demand-withdraw). These are AI opinions anchored to named research concepts, and the report says plainly: the council heard your side only.
| Who | Medicine — MedQA (USMLE-style) | Law — LegalBench (legal reasoning) |
|---|---|---|
| Expert physicians, same questionshuman reference | 87% Liévin et al. ↗ | |
| Human passing threshold (USMLE-style)human reference | ≈60% PLOS Digit. Health ↗ | |
| Passing-lawyer threshold (bar exam MBE — a different test, shown for context)human reference | ≈60% NCBE ↗ | |
| Claude Opus 4.1Anthropic · chairs the consensus | 92.5% vals.ai ↗ | 83.5% vals.ai ↗ |
| Claude Sonnet 4.6Anthropic · senior seat | 92.1% vals.ai ↗ | 82.1% vals.ai ↗ |
| Amazon Nova ProAmazon · document analysis | 81.1% Stanford HELM ↗ | 73.6% Stanford HELM ↗ |
| Claude Haiku 4.5Anthropic · fast seat | 79.6% vals.ai ↗ | 81.2% vals.ai ↗ |
| Llama 4 MaverickMeta · open-weight diversity | 43.3% vals.ai ↗anomalously low vs its own model family on this harness — its legal score is 77.8% | 77.8% vals.ai ↗ |
| Amazon Nova LiteAmazon · fast cross-check seat — cross-checks, never chairs | —no published score for this exact model | —no published score for this exact model |
| Llama 3.3 70BMeta · independent reasoning seat | —closest published sibling (Llama 3.1 70B): 84.8% vals.ai ↗ | 77.2% vals.ai ↗ |
| Mistral Pixtral LargeMistral (EU) · European lab · vision seat | —its text backbone (Mistral Large 24.11): 76.2% vals.ai ↗ | —no published score for this exact model |
2 models — $25 · 5 — $49 · 10 — $99 · 12 (full bench) — $249. Pick the council size below.
Independent published evals of the exact model versions where available — siblings labeled as such, no score invented, re-verified 2026-07-03. Official benchmarks: MedQA paper · MedQA data · LegalBench · LegalBench paper. Independent leaderboards: vals.ai MedQA · vals.ai LegalBench · Stanford HELM · NCBE MBE. Scores move as models ship.
| Couples-therapy course (12–20 sessions) | $2,400 – $5,000 |
| Marriage counsellor, per session | $150 – $300 |
| Couples retreat / intensive | $3,500 – $10,000 |
| MarriageCounsellingCheck AI Opinion & Strategy | $25 – $249 |
The council names the pattern, the likely cause, and step 1 — before you decide whether therapy, a conversation, or a change of your own comes first.
Our system doesn't replace any of them — but it sure helps to educate you before you pay them.
Pay first, then you'll describe your situation, tell the council your country and state/province, and upload documents on the next page — because we store nothing, your details are only ever handled in memory while the council debates, then discarded. The opinion is grounded in your jurisdiction's laws and guidelines, never a default. See a sample report first →
No. It is a 100% AI-generated opinion on your situation — AI models debating the evidence you provided against published couples research. No therapist reviews it. Many people bring the PDF to a counsellor as a starting map.
Yes, and the council is instructed to say so and to name destructive patterns on BOTH sides, including yours. Honest over comfortable is the standing instruction.
A PDF: what the council thinks is wrong, the possible causes with the research concept behind each, where it heads if nothing changes, and a numbered stepwise plan — each step with what success looks like.