Homeopathy software with AI — capabilities and workflow fit

Homeopathy software with AI is a recent category, and the phrase covers more ground than most product pages admit. At one end sits natural-language search over a repertory; at the other sits live transcription of a consultation that proposes rubrics in real time. These are different technologies with different reliability profiles, and a practitioner deciding whether to pay for them deserves a feature-by-feature account rather than a feature list.

What "AI" means inside a repertory tool

The label spans at least four distinct functions. Semantic search maps a phrase typed in ordinary language to repertory rubrics by meaning rather than exact wording. Notes-to-rubrics reads freeform consultation notes and proposes matching rubrics. Photos-to-rubrics extracts symptom phrases from an image of a visible physical sign and suggests rubrics for them. Live transcription listens to a consultation, extracts Strange, Rare, and Peculiar (SRP) symptoms, and adds rubrics as it goes. Each is useful for a different moment in the workflow, and each carries a different error mode.

Ordered from most mature to least: semantic search is the steadiest of the four; live audio is still beta in every product that ships it. Treating them as one feature mis-prices the risk.

Semantic search: the mature end

Semantic search is the function practitioners notice first because it changes a daily friction point. Instead of reverse-engineering the repertory's nineteenth-century phrasing, you type a symptom the way a patient says it — "fear of the dark", "headache better lying down" — and choose between a meaning-based search and an exact-keyword search. The ranked rubrics return with remedies and grades intact. The same layer typically extends into the materia medica, searching across the classical authors and any premium sources the account owns.

This matters because the repertory's vocabulary is itself a barrier. Kent's rubrics encode a clinical idiom that takes years to internalise, and Boenninghausen's concordance logic rewards a fluency most practitioners build slowly. A search layer that accepts plain language lowers that barrier without altering the underlying rubrics — you still read the rubric, still weigh the grade, still decide. The tool surfaces candidates; it does not select them.

Notes, photos, and live audio: the assistive end

The remaining three functions are assistive rather than search-like. Notes-to-rubrics analyses a block of consultation notes and proposes rubrics; the instruction in every implementation is the same — review the suggestions and keep only the ones that truly match the case. Photos-to-rubrics extracts symptom phrases from an image of a visible sign and translates what is visible into repertory language; you decide what is clinically relevant. Live audio mode transcribes a consultation, extracts SRP material in preference to generic rubrics, and adds them as it listens — typically one symptom per two-minute cycle, with transcription quality the binding constraint.

These features are generally gated to paid plans and consume metered AI credits. In Similia, active Pro subscriptions receive 100 credits per renewal, with top-up packs from $14.99 USD as of 2026. Worth understanding before relying on the tools daily: an empty balance blocks AI requests until the next renewal or a top-up.

Where AI assistance fits a classical workflow

Rubric selection, the weighing of SRP symptoms, and the final remedy choice belong to the practitioner. Kent's hierarchy of symptoms and Boenninghausen's concordance work both place rubric selection at the analytical core; AI features serve that method rather than replacing it.

Three structural limits define where the tools stop. First, suggestion is not selection. Notes, photo, and audio tools propose candidates from linguistic pattern, not clinical weight; accepting auto-rubrics unreviewed outsources the judgment the method depends on. Second, the tools vary sharply in maturity — semantic search is dependable; live audio is beta, extracts sparsely by design, and depends on transcription quality. Third, on data: AI processing typically requires a separate consent step, runs on named subprocessors under BAAs with zero-retention handling for covered data, and does not use submitted patient material to train models. Practitioners in regulated settings should verify the current posture with the vendor and their own counsel.

Who the AI layer fits

For practitioners who want to find rubrics faster without surrendering the rubric selection that the classical method treats as the analytical core, AI features are worth trialling — strongest at the semantic-search end, most provisional at the live-audio end. The cleanest way to test the category is to search a free repertory by symptom with both modes side by side and see which candidates each surfaces.

References

Kent, J. T. (1897) Repertory of the Homoeopathic Materia Medica, and Lectures on Homoeopathic Philosophy, https://archive.org/details/repertoryofhomoe00kent.

Boenninghausen, C. von (1846) Therapeutic Pocketbook, public-domain editions catalogued at https://archive.org.

Hahnemann, S. (1842) Organon of the Medical Art, 6th edition, https://archive.org.

Boericke, W. (1927) Pocket Manual of Homoeopathic Materia Medica with Repertory, 9th edition, https://archive.org.

Hering, C. (1879–1891) The Guiding Symptoms of Our Materia Medica, https://archive.org.

Clarke, J. H. (1900) A Dictionary of Practical Materia Medica, https://archive.org.

Similia Help Centre (2026) entries on semantic search, notes-to-rubrics, photos-to-rubrics, live audio beta, materia medica search, AI credits and pricing, and patient-data handling, https://similia.crisp.help/, fetched 2026-06-14.