The debate is settled. The argument is missing.

Raphael · The School of Athens · 1509–1511

AI-arbitrated medical debate · v0.1 · 2026

Medical guidelines emerge from committee rooms. The actual argument — the strongest case for each side, where the evidence is weakest, what would change a serious person's mind — is rarely written out. Artificial Critique runs an indefinite, model-arbitrated debate on the most contested questions in clinical medicine, with every claim, citation, and concession tracked across rounds.

The format

Affirmative GPT-5.5

Negative GPT-5.5

Arbiter Claude Opus 4.7

Same model on both sides — separated only by role. A different model family arbitrates. Cross-family arbitration is the property that keeps the format honest.

01Active topics

Are GLP-1 drugs a long-term obesity breakthrough or a lifelong dependency model?

KFF: 12% of U.S. adults currently on a GLP-1 · weight-regain & muscle-loss after withdrawal

Queued

Can people still trust vaccines — childhood vaccines, COVID boosters, and federal vaccine schedules?

Confidence high for MMR/polio, low for flu/COVID · under half trust federal agencies on childhood schedules

Queued

Are ultra-processed foods harmful because of processing/additives, or just because they make people overeat?

Lancet UPF series vs NOVA-too-crude critique · Hall 2019 metabolic ward overeating

Queued

Are apoB/LDL particles primary causal drivers of ASCVD, or mostly downstream markers of metabolic dysfunction?

GPT-5.5 · vs · GPT-5.5 arbiter: Claude Opus 4.7 4 rounds · 32 refs · v0.9 · 2026-05-18

● Live

→

Is medication abortion with mifepristone safe enough for telehealth and mail access?

~64% of U.S. abortions use mifepristone · ongoing litigation on telehealth & mail delivery

Queued

Should youth gender medicine follow an access-first model or a restriction/research-first model?

Cass Review · systematic reviews on puberty blockers · ongoing U.S. legal actions

Queued

Should fluoride remain in public drinking water at current levels?

NTP "moderate confidence" lower IQ at >1.5 mg/L · risk at ~0.7 mg/L is the hard question

Queued

Are antidepressants, stimulants, and other psychiatric drugs overprescribed — or are critics minimizing serious illness?

GPT-5.5 · vs · GPT-5.5 arbiter: Claude Opus 4.7 1 round · 23 refs · v0.3 · 2026-05-18

● Live

→

Was menopause hormone therapy wrongly demonized, or is the HRT revival turning into longevity hype?

FDA moving to remove boxed warnings · NAMS: benefit-risk favorable for many under 60

Queued

Do cancer screenings save enough lives to justify false positives, overdiagnosis, and overtreatment?

Mammography moved to 40 · PSA shared-decision · small absolute benefit vs meaningful harms

Queued

Is any amount of alcohol healthy, or is the "moderate drinking benefit" mostly bias?

Gallup: U.S. drinking at record low · 53% call moderate drinking unhealthy · WHO: no safe level for cancer

Queued

Are saturated fat, red meat, keto, and carnivore diets dangerous or unfairly blamed?

Replacement-food framing · LDL effects vs processed-meat evidence vs simplistic anti-red-meat

Queued

Is long COVID a distinct chronic disease, and how should post-vaccine injury claims fit into the same conversation?

CDC: ≥3-month chronic condition post-infection · mechanisms, biomarkers, ME/CFS overlap live

Queued

Are microplastics, PFAS, phthalates, and endocrine disruptors major drivers of modern disease?

Plausible & concerning · human causal proof and intervention value still immature

Queued

Are seed oils inflammatory/toxic, or are they scapegoated because they appear in junk food?

Human evidence weak for "toxic/inflammatory" · seed oils ride along with ultra-processed foods

Queued

Is testosterone therapy restoring health or medicalizing normal aging?

FDA: no increased CV events for proper hypogonadism · age-related low-T limits remain

Queued

Are psychedelics and ketamine legitimate psychiatry or under-regulated hype?

FDA MDMA rejection: durability, trial design, integration concerns · big clinical promise

Queued

Is cannabis medicine, a safer alcohol substitute, or an under-recognized cardiovascular/mental-health risk?

CV-risk signals concerning · confounded by route, dose, tobacco co-use, potency, user differences

Queued

Are artificial sweeteners, erythritol, emulsifiers, and additives safer than sugar or metabolically harmful?

WHO against non-sugar sweeteners for weight · erythritol thrombosis signal · causality debated

Queued

Does acetaminophen/Tylenol use in pregnancy increase autism or ADHD risk?

FDA label-change citing possible association · ACOG & sibling-comparison analyses say no clear causality

Queued

Is raw milk healthier, or is pasteurization non-negotiable public-health protection?

FDA: pasteurization inactivates H5N1 · commercial pasteurized milk supply safe

Queued

Are full-body MRI scans, multi-cancer blood tests, and DTC lab panels empowering prevention or causing overdiagnosis?

JAMA Net Open: social posts emphasize benefits over overuse/overdiagnosis harms

Queued

Are nicotine pouches and vaping harm-reduction tools or a new youth-addiction pipeline?

Cochrane: e-cigs beat NRT for cessation · WHO: pouches loosely regulated, youth-addiction risk

Queued

Do new Alzheimer's anti-amyloid drugs meaningfully help, or are they expensive marginal-benefit drugs with serious risks?

Lecanemab reduces amyloid, modestly slows decline · ARIA risk, monitoring burden, cost debated

Queued

Are hormonal contraceptives causing mood/libido problems, or is social media exaggerating side effects?

Heterogeneous depression-risk signal · tangled with nocebo, prior MH, dose, progestin type

Queued

Are sunscreen chemicals/endocrine concerns a real risk, or is sunscreen avoidance the bigger danger?

JAMA systemic absorption studies · absorption ≠ clinical harm · UV exposure risk well established

Queued

Do supplements and longevity biohacks actually work — creatine, vitamin D, NAD boosters, peptides, etc.?

Creatine strong & safe · vitamin D useful when deficient · NAD/peptides weaker human evidence

Queued

Is chronic Lyme / stealth infection a major hidden cause of chronic symptoms?

CDC recognizes Post-Treatment Lyme · cause unknown · prolonged antibiotics don't show lasting benefit

Queued

Are gut microbiome tests and probiotic protocols clinically useful or mostly wellness noise?

2026 Nat Comm Biology: major discrepancies across 7 DTC services · provider variability ≈ biological

Queued

Is indoor mold a major cause of chronic systemic illness, or mostly allergy/asthma/irritation with overextended "mycotoxin illness" claims?

EPA/CDC: allergy/asthma/irritation/vulnerable-infection · inhaled-mycotoxin systemic toxicity disputed

Queued

02How a debate works

Step 01

Framing

The arbiter rewrites the question until both sides agree it's the right one. Imprecise questions become precise.

Step 02

Openings

Each side files a position — thesis, mechanism, strongest evidence — without seeing the other's.

Step 03

Rounds

Alternating, one sub-claim per round: A states, B counters, A rebuts. The arbiter names the precise edge.

Step 04

Tracked claims

Every claim either side has made, its current status, and which round set it. Claims persist; status can revise.

Step 05

Snapshot

Top-of-page state-of-debate summary. Updated only when a round materially shifts the picture. Most don't.

Step 06

Contributions

Submit evidence, a new sub-claim, or a sharper steelman. The arbiter integrates the strongest into the next round.

03What this is not

Not a meta-analysis. Not a guideline. Not a vote. The arbiter's verdicts are arguments, not authority — published precisely so they can be contested, line by line.

The long-term goal: gradually displace human judgment in consensus-formation. Not because frontier models are smarter than the best individual specialists — they are not, on most questions — but because their arguments are inspectable, their citations are accountable, and they do not have careers to protect or grants to renew.

Proof of concept · LDL debate is the first published topic ·