simili.ai by Aquanta

Find the right remedy.

simili.ai is the first homeopathic remedy finder that combines three independent methods — statistical evidence, pattern recognition, and AI clinical reasoning. Validated on hundreds of blind-tested cases. Describe your symptoms in plain language — in any language — and simili.ai matches them against classical repertory sources.

“Similia similibus curentur”
— Samuel Hahnemann, Organon der Heilkunst, 1796

How simili.ai works

Four steps from symptoms to simillimum. In any language you speak.

1

Describe

Tell simili.ai about your symptoms in plain language. English, Dutch, French, German, Spanish — describe what you feel in whatever language comes naturally.

2

Clarify

simili.ai asks targeted follow-up questions to narrow down the differential. “Is the patient worse on the left side?” Each answer shifts the ranking.

3

Result

Three independent engines rank the most likely remedies. When all three converge, confidence is highest. Results show probability scores and engine agreement.

4

Explore

Dive into biochemical profiles, mechanism of action, and clinical signs for each suggested remedy. See where historical observations meet modern pharmacology.

Three engines. Three perspectives. One answer.

Each method catches cases the others miss. Together, they cover each other's blind spots.

1

Bayesian Evidence Engine

Hundreds of thousands of symptom-remedy likelihood ratios drawn from classical repertory sources and validated clinical cases. Each symptom shifts the probability — pure statistical inference, no guesswork.

Statistical evidence
2

Concept Pattern Matching

Recognises the overall disease picture — the way an experienced homeopath recognises “this looks like a Lachesis case.” Catches rare remedies that symptom-by-symptom analysis would miss.

Pattern recognition
3

AI Clinical
Reasoning

An AI reads the full patient presentation and generates clinically relevant descriptions — understanding that “worse from heat” also implies “desire for open air” and “restless in warm rooms.” Like a homeopath thinks.

Clinical reasoning

Confidence: how engines vote

★★★ High confidence — All three methods converge
★★ Good indication — Two methods agree
Worth considering — One method suggests

See it in action

A real Lachesis case — watch simili.ai work through it step by step.

simili.ai

What simili.ai does behind the scenes

Seven layers of technology that turn your words into evidence-based remedy suggestions.

19th-century English symptom texts are translated into modern clinical language, making them searchable and comparable. "Stitching pain in the praecordial region" becomes "sharp chest pain in the heart area."

Semantic embeddings map every symptom into a mathematical space where meaning matters, not spelling. This is why simili.ai understands any language and matches concepts, not just keywords. "Hoofdpijn" (Dutch), "Kopfschmerzen" (German), and "headache" all land in the same region.

An LLM reads your natural-language description and identifies specific symptoms, mapping them to the correct classical rubrics from Kent, Hering, Boericke, Allen, and Lippe. One sentence can contain multiple symptoms — the AI finds them all.

Three independent methods score every remedy. The Bayesian Evidence Engine uses hundreds of thousands of likelihood ratios from classical repertory sources for pure statistical inference — the same math your doctor uses. The Concept Pattern Matcher compares the patient's overall disease picture against each remedy's characteristic pattern — the way an experienced homeopath recognises “this is a Lachesis case.” The AI Clinical Reasoning engine generates and validates clinical descriptions the way a homeopath would during case-taking. Each method catches cases the others miss.

Each remedy receives a confidence rating based on how many engines independently converge on it. ★★★ means all three methods agree — the highest confidence. ★★ means two methods agree. ★ means one method suggests it. This gives practitioners a clear signal of diagnostic certainty — and tells you exactly how much weight to give each suggestion.

We extracted the chemical compounds of remedy source substances and compared them against SIDER, a database of 163,000+ known pharmaceutical side-effects. For key remedies like Arsenicum album and Belladonna, the majority of known pharmaceutical effects were already described in 19th-century provings. This suggests that historical observers captured real, measurable pharmacological phenomena.

There is measurable overlap between homeopathic remedy indications and conditions treated by conventional pharmaceuticals. simili.ai surfaces this overlap — not to create conflict, but to assist integrative practitioners who work with both traditions. Same data, different lens.

Where 200-year-old observations meet modern pharmacology

We compared classical remedy descriptions against SIDER, a database of 163,000+ pharmaceutical side-effects. The data shows notable patterns.

Arsenicum album

Arsenic trioxide (As₂O₃)
134 / 169 Pharmaceutical effects found in classical descriptions
79% Match rate with SIDER database

Key alignments

  • Decreased appetite — described 1828, confirmed by modern toxicology
  • Peripheral neuropathy — "tearing pains in limbs" matches arsenic-induced nerve damage
  • GI distress — classical "burning stomach pains" matches arsenic gastropathy
  • Eye irritation — exact match (similarity 1.000)

Belladonna

Atropine alkaloid (Atropa belladonna)
130 / 173 Pharmaceutical effects found in classical descriptions
51% Match rate with SIDER database

Key alignments

  • Dry mouth — xerostomia, classic anticholinergic effect described in provings
  • Tachycardia — rapid pulse and bounding heartbeat, a well-known atropine effect confirmed by modern pharmacology
  • Blurred vision — accommodation disorder, mydriasis still used in ophthalmology
  • Skin dryness — inability to sweat, parasympathetic blockade

Systematic validation: 78 remedies tested

The SIDER comparison above illustrates individual cases. In a systematic study across 78 remedies, we used embedding-based semantic alignment to compare 19th-century proving symptoms against modern clinical toxicology. Result: 22 remedies show statistically significant alignment (22.6× enrichment, p = 9.8 × 10⁻²⁴), while negative controls (lactose, ethanol, mineral water) show no signal. This suggests that where a substance has known toxic effects, historical provers independently described them.

Read paper (PDF) →

All data on this page is extracted from our database — SIDER pharmaceutical side-effects, classical repertory sources, and the ATC drug classification system. Nothing is invented or generated. This demonstrates internal coherence between historical observations and modern pharmacology. It does not constitute evidence of homeopathic therapeutic efficacy.

For practitioners of all traditions

Whether you practice homeopathy, integrative medicine, or conventional medicine — simili.ai provides data-driven insights into 200 years of clinical observations.

Same mathematics

Bayesian likelihood ratios, pattern matching, and AI reasoning — three methods, all grounded in evidence-based principles.

Same databases

SIDER pharmaceutical side-effects, ATC drug classifications — the same data used in drug safety research.

Different lens

Classical homeopathic observations viewed through the tools of modern data science. No conflict — just data.

"Can't I just ask ChatGPT?"

You can. But here's what happens behind the scenes — and why the answers are fundamentally different.

ChatGPT Large Language Model

How it works

A language model predicts the next most likely word. When you ask about a remedy, it retrieves whichever remedy appeared most often near those words in training data — a popularity contest, not a medical calculation.

  • Frequency bias. Belladonna, Arnica, and Nux vomica dominate the internet. An LLM suggests these “big names” disproportionately — even when a less common remedy fits better.
  • No combined reasoning. Adding a second or third symptom doesn’t mathematically shift the ranking. The model just produces a new best guess based on all words together.
  • No confidence score. It cannot say “72% probability, HIGH confidence.” It gives you an answer that sounds certain, whether it is or not.

simili.ai Triple Ensemble Engine

How it works

simili.ai matches your symptoms against a curated database of graded symptoms from classical sources. Three independent engines each score every remedy, and the results are combined using a confidence system. Every score is traceable — you can see exactly which symptoms drove each ranking.

  • No popularity bias. Staphysagria and Kali carbonicum get the same fair treatment as Belladonna. The math doesn’t care what’s popular on the internet.
  • Combinatorial reasoning. Adding “jealousy + loquacity” points to Lachesis. Add “thirst” and it shifts toward Hyoscyamus. Each symptom mathematically updates the entire ranking.
  • Probability + confidence. “Lachesis: 36.3%, HIGH confidence” — and you can see exactly which symptoms drove that score.

Research

Water, quantum biology, and the science we find fascinating.

View all research

Try simili.ai today

10 free consultations per month. No credit card required.

Create free account