How the AI works
From an entry to an interpretation.
How Get Regular uses meal, stool, symptom and chat entries for AI-assisted estimates and insights, and how to check an interpretation against your original record.
AI can help organize information you enter. Understanding where an observation ends and an interpretation begins is essential to using the output sensibly.
First, the information you supply
Get Regular’s public product description includes photo or text meal records, stool tracking, symptoms and Gut Check conversations. The privacy policy identifies OpenAI for AI analysis of selected health content and chat messages. These are the information categories behind the workflows described here.
An entry can be incomplete before AI sees it. A photo may not show an ingredient, and a remembered symptom may not have a precise start time. An analysis cannot recover missing facts with certainty. That uncertainty should remain visible when interpreting a response.
| Your input | Use described by the product | Keep in mind |
|---|---|---|
| Meal photo or text | Ingredient and nutrition estimates | A photo may miss ingredients, preparation details and portions. |
| Stool photo or manual Bristol entry | Stool-form recording and analysis | A classification describes appearance; it does not establish a cause. |
| Symptoms and diary context | Possible patterns across recorded history | Missing entries are unknown information, not symptom-free days. |
| Gut Check conversation | Responses personalized with logged information and stored conversational facts | Check the response against what you actually recorded. |
Then, several different kinds of output
A description of a photo, an estimated nutrition breakdown, a numerical score and a conversational response are different outputs. They should not be treated as interchangeable measures of health. A well-written answer is not a measurement, and a number does not become a validated measurement because it appears in a dashboard.
AI can make errors in an ingredient, date, stool description or explanation. Read the output as something to inspect alongside your record. The public feature descriptions do not establish a clinical accuracy percentage or show that every response has been checked by a clinician.
| Information | How to read it |
|---|---|
| Your entry | Your recorded observation; its detail and accuracy depend on what you supplied. |
| AI estimate or classification | An interpretation that may miss context or need correction. |
| Generated suggestion | A possible discussion point, not an established cause or treatment. |
Follow the job from input to output
The public descriptions outline three jobs: estimate what a meal or stool photo shows, summarize recorded history, and answer questions in Gut Check. A photo estimate, Weekly Digest, score and chat response are different kinds of output; an accurate description in one task does not prove accuracy in another.
Manual entry lets you record an observation without uploading an image. It does not mean that the resulting record stays only on your device or bypasses all AI processing. Read the privacy explanation before choosing which health details to enter.
This explanation concerns the diary and conversation inputs described by the product. It does not describe a laboratory test, a blood-test interpretation service or a substitute for clinical assessment. An Apple Health connection also should not be assumed to supply every kind of information to AI.
Check the interpretation against the record
Compare the response with the entries it describes. Ask whether it names information you never supplied, overlooks a gap or turns a sequence into a cause. If the explanation cannot be checked, keep that limitation rather than filling the gap with confidence.
Gut Check does not replace a clinical assessment. Questions about a diagnosis, medication, significant dietary restriction or whether to seek care belong with an appropriate healthcare professional. Do not rely on a reassuring app response to rule out a health concern.
- A hidden ingredient is missing information, not something a photo proves was absent.
- A repeated food before symptoms is an association, not an isolated cause.
- A polished answer can contain the wrong entry or timeline; compare it with the record.
- A missing log is unknown information, even when the output is a complete-looking summary.
Your questions, answered.
Is an AI interpretation the same as my log?
No. Your log records what you entered. An AI estimate, summary or response interprets some of that information and can introduce mistakes. Check the two together.
Does manual entry keep all processing on my phone?
Manual entry avoids uploading a photo for that observation. The privacy policy describes cloud processing of health entries and chat content, so manual entry should not be treated as an on-device-only promise.
Does every answer prove a food trigger?
No. Timing and repeated associations are observations to examine. They do not establish an intolerance, allergy or cause of symptoms.