Kerith

AI Disclosure

Last updated: October 20, 2026

Provided in accordance with Apple App Review Guideline 5.1.2(i).

How Kerith uses AI

Kerith uses third-party AI models to power its interactive features. By default, text generation — Bible study chat, devotionals, prayer reflections, journal reviews, sermon summaries, reading tracks, and verse-memorization aids — is performed by OpenAI's GPT models. When OpenAI is temporarily unavailable, the same request is handled by Google's Gemini models instead, so the feature keeps working. Reflections on your journal entries are generated by Google's Gemini models. Sermon photos you upload are processed by Google Gemini to extract their text. Sermon audio you submit is transcribed by OpenAI's Whisper or by Cloudflare Workers AI. Semantic search and personalization use text embeddings generated via Voyage AI and Cloudflare.

Requests to OpenAI and Google are routed through Cloudflare AI Gateway. The gateway does not store the content of your requests or of the AI's responses; it keeps a short-lived record of request metadata (model, token counts, cost, and a pseudonymous identifier that is not your name, email, or account ID).

We may change or add AI providers over time; the current list is maintained in our Subprocessors list. AI features are opt-in: no content is sent to any AI provider until you grant AI consent in the app (see "Your control" below).

What data is sent to AI providers

The following data may be sent to third-party AI providers when you use specific features:

Your name, email address, and Apple ID are never shared with AI providers. AI providers process your content under their own terms, which do not permit using your content to train their models. See our Privacy Policy for full details.

How responses are generated

Each AI response is generated for your specific query. We do not reuse one user's responses for another user. We apply theological guardrails in our system design to avoid taking sides on debated denominational questions and to encourage consulting pastoral authority for serious matters. The AI is instructed to direct you to your pastor or church community when questions exceed the scope of study assistance, and to surface crisis-support information when a message suggests self-harm.

Limitations

AI can and does make mistakes. Scripture references generated by the AI should be independently verified. Theological interpretations may reflect the AI's training data and may not align with every denomination's position. If you receive a response that seems theologically incorrect or harmful, please report it using the in-app feedback button so we can investigate and improve.

Your control

You have meaningful control over AI processing in Kerith: