Data Usage

How a Ripenn audit works

Each visibility check is a two-step pipeline:

1. Engine call. Ripenn sends a prompt from your prompt set to an AI engine (ChatGPT, Claude, Gemini, or Perplexity). The engine answers with text and, where supported, cited sources. The full response is stored in Ripenn's database.
2. Analysis call. A second AI call reads the engine's response and extracts structured data: whether your brand was mentioned, sentiment, sources cited, competitors named.

This means every prompt in your prompt set is sent to at least one external AI provider during an audit cycle.

AI providers used for audit runs

The providers below receive the prompts in your prompt set when audits run. They do not receive your account information, your billing details, or content from other parts of Ripenn.

Provider policies change. The summary above reflects each provider's current public terms; the authoritative source is each provider's own data-usage documentation, which Ripenn re-checks on a quarterly cadence.

What we do not send to AI providers:

• Your account credentials, billing information, or Ripenn user list
• Audit results from other customers
• Any data outside the specific prompt being scored on that run

Analysis calls

The analysis step uses one AI provider (currently Anthropic Claude) to classify the engine's response. The analysis call receives:

• The engine's response text from step 1
• The brand name and competitor names you configured
• Instructions for the classification task

Analysis providers are subject to the same training and retention rules as their corresponding engine providers above.

What Ripenn stores from an audit

For each check, Ripenn stores in its Supabase database:

• The prompt text
• The engine's full response (text plus any cited URLs)
• The structured analysis output (mentioned / not mentioned, sentiment, sources, competitors)
• Metadata: which engine, which run number, timestamp, token counts, cost

These records live in your project and are accessible only to authorized members of your workspace, enforced by row-level security. See Security for the multi-tenant isolation details.

How Ripenn uses customer data internally

Beyond running the service for you, Ripenn uses customer data in one narrow way: aggregate analytics across all customers.

This means counts, percentages, averages, and statistical patterns computed across the full Ripenn dataset — never tied back to any individual customer, brand, or prompt. Examples of what aggregate analysis looks like in practice are published openly on our research blog:

• Mention-rate variance across industries
• The relationship between number of runs and measurement accuracy
• Cost-per-check breakdowns across engines
• Coverage tradeoffs in query set design

When we publish this research, industries are described generically (“loyalty & promotions”, “no-code software”, “tree care retail”) and no customer is identified. Individual prompts, brand names, customer domains, and audit responses do not appear in published research.

What Ripenn does not do with customer data

These are explicit commitments:

• Ripenn does not train foundation models on customer data. We do not operate any model training pipeline that ingests customer audit responses, prompts, or content.
• Ripenn does not share customer audit data with third parties for their training. When we call OpenAI, Anthropic, Google, or Perplexity to run an audit, each call is subject to that provider's default API terms (which exclude training, as summarized above).
• Ripenn does not sell customer data. Not aggregated, not pseudonymized, not in any form.
• Ripenn does not use one customer's data to benefit another customer in a way that would surface it. A competitor analysis run for Customer A is not visible to Customer B, and Customer A's prompt set, content, or audit responses are never used to populate Customer B's project.
• Ripenn does not look at individual customer projects for product-improvement purposes outside of the aggregate analysis described above, except when responding to a specific customer support request from that customer.

How long audit data is kept

• Active audit responses are retained for the lifetime of the project.
• Deleted projects are cascade-deleted from the database. Backup copies (see Security) age out within the 7-day backup window.
• Provider-side retention for prompts sent to AI engines is governed by each provider's own policy, summarized in the table above.

Opting out

Customers may opt out of having their data included in Ripenn's aggregate analysis (the basis of the published research described above) by emailing support@ripenn.ai. Operational uses required to deliver the service — running your audits, storing your results, billing — cannot be opted out of without terminating the account.

Provider changes

Ripenn periodically swaps AI providers or model versions as the underlying landscape changes (a new GPT version, a new Claude release, a switch from one Gemini tier to another). When the category of provider changes — for example, adding a new AI engine to the audit pipeline — this page is updated and existing customers are notified by email. Minor model-version changes within an existing provider are noted on the changelog but do not trigger a notification.

Questions

For specific questions about how Ripenn handles your data, email support@ripenn.ai.