How Iridology Works: Our Methodology
A plain walkthrough of what happens between your two photos and your reading: the quality check, the model, the fourteen-system chart and the four personality types. We say what the system does and where its limits lie.
The Reading Pipeline
Five stages from two photos to a finished reading
When you upload a photo of each eye, they pass through five stages. Each has one job and hands its output to the next. The whole run takes about a minute. If you have not scanned yet, the photo guide shows how to take the two photos.
1. Two Photos and a Quality Check
You take one photo of each eye with your phone's back camera, standing at a mirror with the phone touching your face. The back camera has the sharper lens; the front camera is too soft to show the fibres. Both photos are then scored for how clearly the iris is visible. If either eye scores below the threshold you are told which one and why, so you can retake it.
This step matters because everything downstream assumes a clear iris. A soft or half-covered photo does not produce a vague reading, it produces a confident-sounding wrong one, which is worse. The gate is conservative on purpose.
2. Reading the Iris
The two photos are passed to a multimodal AI model together with a long, fixed set of instructions. The instructions name the chart to read against, the fourteen body systems to report on, and the rule that every finding must cite something actually visible: a lacuna at a clock position, a contraction ring, pigment at the collarette, a difference between the left and right eye.
The model writes the reading in structured form: a score and a finding for each system, strengths, areas to nurture and suggestions. Our software checks that the structure is complete and rejects a response that is not.
3. The Fourteen-System Chart
The Health reading reports on fourteen systems: respiratory, digestive, endocrine, reproductive, muscular, nervous, lymphatic, skeletal, urinary, cardiovascular, circulatory, excretory, immune and integumentary. Each is tied to the region of the chart tradition assigns it, and the reading says which region it read.
The chart follows one well-documented convention throughout. That keeps every reading on the same map; it does not make the chart a standard, because iridology has none.
4. The Four Personality Types
The Personality reading uses behavioural iridology, sometimes called the Rayid method, which reads the structure of the iris rather than its health. It names four types: Jewel (pigment spots, the thinker), Flower (rounded open lacunae, the feeler), Stream (fine straight fibres, the grounded) and Shaker (a mix of jewels and flowers, the driver). Most irises blend two.
From the dominant and secondary type, plus fibre density, contraction rings and the difference between eyes, the model writes an archetype, five trait scales, strengths, blindspots and a short development plan. It is a reflective lens, like a well-made personality quiz, not a scientific test.
5. Laying Out the Reading
The structured result becomes the pages you see: the vitality index, the systems mandala, the annotated body map, and for the Complete reading, the chapters that read the two halves against each other. The paid readings are also typeset as a PDF.
We do not claim the reading has been validated against clinical outcomes. It describes what the iridology tradition says about the patterns it can see, and the same photos can give a somewhat different reading on another run. Whether those patterns reflect underlying physiology is a separate question, and we return to it below.
Research Basis and Limitations
What the evidence supports and where it falls short
What the Model Can Do
Reading fine detail in a photograph, describing it in plain words and following a long set of rules are things current vision-language models do well. That is the part of the pipeline we rely on: every finding must point at a feature the model can see. The model writes each reading fresh, so the same photos can come back with somewhat different wording and scores.
What the Research Shows About Iridology
The evidence for iridology itself, the idea that iris patterns systematically reflect the state of internal organs, is mixed at best. A 2005 systematic review in the Journal of General Internal Medicine examined the controlled trials available and concluded iridology was not a useful diagnostic tool. Some later work reports more favourable findings, but the body of evidence remains contested and no study validates it as a diagnostic method.
We are plain about this: the research does not support iridology's core diagnostic claims, and we do not make them. The reading works inside the tradition as a practice of self-reflection. The patterns it describes in your iris are real; what they mean for your health is an open question.
Our Approach to Validation
We claim no FDA clearance, CE marking or clinical validation. What we check is that the reading is complete and that every finding cites a visible feature. Clinical accuracy would need outcomes data we do not have, and we would rather say so than imply otherwise.
Data Privacy and Security
What happens to your photos, stage by stage
Iris photos are biometric data and we treat them that way. Here is what actually happens to them:
Transit encryption
Uploads travel over TLS. The photos leave your phone encrypted and stay encrypted on the way to the model and to storage.
Stored with your reading
Both photos are kept privately alongside the reading they produced, so you can reopen it and so a reading can be rewritten from the same photos if the format improves. They are used for nothing else, unless you choose to keep them for research.
Who sees them
To write your reading, the photos are sent to an AI provider: OpenAI's GPT models. They are never sold. Readings sit in your account and are not visible to other users.
Right to deletion
Delete your account from your account page, or email support@iridology.app, and your readings and their photos are deleted. The one exception: photos you chose to keep for research stay, with nothing that links them to you.
Our privacy policy has the legal detail. The short version: your photos go to the AI provider that writes your reading, stay private with it, and are deleted with your account unless you chose to keep them for research. We do not build a biometric database and we do not sell them.
Open Questions and Future Directions
What we are still working on
Several parts of the method are still moving:
Multi-session tracking
Comparing readings taken months apart, so you can see whether a system's score moved. Different phones, lighting and distances all confound that comparison, and we would rather solve it properly than show a misleading trend.
Calibration
Scores follow one fixed scale but are not calibrated against any external benchmark, and they can shift a little between runs on the same photos. We are looking at ways to make them more interpretable and more comparable between people.
Better evidence in every finding
Each finding already has to cite a visible feature. We want the reading to show that feature on your own photo, so you can check it yourself.
Independent validation
We would welcome researchers outside the team evaluating the system. Any tool that makes health-adjacent observations should be open to that scrutiny.
These are not promises. Some may ship soon, some may take longer, some may not pan out. We will update this page as the method changes.
Frequently Asked Questions
What kind of AI does Iridology use?
A large multimodal model that can look at images and write text. We send it both iris photographs with a detailed set of instructions: the iridology chart it should read against, the fourteen body systems it must report on, the four behavioural iridology types it may name, and the rules for saying nothing it cannot see. It returns a structured reading that our software checks and lays out. We do not train our own network on iris images.
How does the quality check work?
Before any reading is written, a cheaper pass looks at each photo and scores how clearly the iris is visible from 0 to 100. Both eyes must score at least 50. Blur, a closed lid, heavy reflections or a photo that is not of an eye all fail the gate, and you are asked to retake. The gate is deliberately strict: it turns away some usable photos so that fewer poor ones produce a misleading reading.
Is Iridology FDA cleared or CE marked?
No. Iridology is a tool for self-reflection and has no FDA clearance, CE marking or medical device certification. It reads iris patterns according to iridology chart conventions and writes a wellness-oriented reading. It does not diagnose, treat or prevent any disease. Treat the reading as a complementary observation alongside professional medical advice.
What happens to my iris photos after the scan?
Both photos are sent over an encrypted connection to an AI provider, OpenAI's GPT models, which writes your reading. They are then stored privately with your reading so you can reopen it. They are never sold. When you delete your account they are deleted too, unless you chose to keep them for research. Our privacy policy has the legal detail.
How does zone mapping work in practice?
Iridology charts divide the iris into regions, by clock position and ring, that tradition assigns to specific organs and systems. The model is instructed to read against one widely referenced chart and to report on the same fourteen systems every time, naming the clock position and the eye where it sees each feature. We acknowledge that different iridology schools place organs slightly differently. Following one convention consistently makes the reading internally consistent, not universal.
Ready to get started?
Two phone photos, one of each eye, and your reading arrives in about a minute.
Scan my eyes — free