AI Transparency Policy

Our Commitment to AI Transparency

SmartEnneagram is an AI-powered service. Artificial intelligence is not a peripheral feature - it is central to how our personality assessment works. We believe you have the right to understand exactly how AI is involved in analyzing your personality and generating your results.

This Transparency Policy explains, in plain language, which AI models we use, what data they receive, how they process your responses, what limitations they have, and what we do to ensure responsible AI practices. We update this policy whenever we make material changes to our AI systems.

AI Models We Use

SmartEnneagram uses three independent AI language models to analyze your assessment responses. Using multiple models provides diverse analytical perspectives and reduces the risk of any single model's biases dominating your results.

How AI Analyzes Your Responses

When you complete our Enneagram assessment, here is exactly what happens at each stage:

Data Flow - What AI Providers Receive

Transparency about data sharing is essential. Here is precisely what data is and is not shared with our AI providers:

Data Sent to AI Providers

• Your assessment responses (the answers you select or type for each question)
• The question text itself (so the AI can understand the context of your answer)
• Enneagram framework context from our prompt engineering (scholarly references, type descriptions, analytical frameworks)
• A session identifier (a random, non-identifying string used only to maintain processing continuity)

Data NOT Sent to AI Providers

• Your name, email address, or any contact information
• Your IP address, device information, or browser fingerprint
• Your payment information or transaction history
• Any data from previous assessment sessions (unless you explicitly choose to include prior results)
• Any data from other users

Algorithmic Transparency

We aim to be transparent about how our assessment produces results, while protecting our proprietary methodology:

What We Disclose

• Scoring approach: Our assessment uses a weighted scoring system where different questions contribute different weights to type scores. The weights are derived from Enneagram scholarship and refined through iterative testing.
• AI role: AI provides qualitative analysis that supplements (but does not replace) the quantitative scoring. AI analysis adds narrative depth and identifies patterns that pure scoring may miss.
• Multi-model rationale: We use three independent models specifically to reduce the risk of any single model's biases affecting your results. When models disagree significantly, this is noted in your report as an area of complexity in your personality profile.
• Adaptive questioning: The premium 65-question adaptive section selects questions based on your initial 50-question responses, focusing on areas where your type is less clear to provide more precise differentiation.

What We Protect

Our specific prompt engineering, scoring weights, question selection algorithms, and AI integration architecture are proprietary trade secrets. We disclose the principles and approach but not the exact implementation details, as this would enable copying and could be used to game the assessment.

Bias Awareness Statement

We acknowledge that AI systems can reflect and amplify biases. SmartEnneagram takes the following approach to AI bias:

Known Bias Risks

• Cultural bias: AI models are primarily trained on English-language, Western-centric data. Enneagram descriptions and our AI prompts may reflect Western cultural assumptions about personality that do not translate universally across cultures.
• Language bias: Users whose first language is not English may receive less nuanced analysis, as AI models process English-language input most effectively.
• Historical bias: The Enneagram literature itself was developed primarily by Western, educated, middle-class authors. Our AI analysis reflects the perspectives present in this body of literature.
• Model training bias: The AI models we use were trained on internet text that may contain stereotypes, biases, and cultural assumptions. Despite provider efforts to reduce bias, these effects cannot be fully eliminated.
• Confirmation bias: AI models may produce analysis that confirms expectations rather than challenging them, a known limitation of language model behavior.

Bias Mitigation Efforts

• Using three independent models from different providers reduces single-source bias
• Our prompts explicitly instruct AI models to avoid stereotyping and to acknowledge individual variation within types
• We periodically review AI outputs across demographic categories to identify systematic bias patterns
• We welcome user feedback about biased or culturally insensitive results at [email protected]

Accuracy Limitations

We are honest about what our AI-powered assessment can and cannot do:

• Not clinically validated: Our assessment has not undergone independent psychometric validation by an accredited research institution
• Not diagnostic: Results do not constitute a psychological diagnosis and should not be treated as such
• Probabilistic: AI models are probabilistic systems. The same input may produce slightly different outputs on different occasions. This is a fundamental characteristic of current AI technology, not a defect.
• Self-report limitations: Like all self-report assessments, our test relies on your self-perception, which may differ from how others experience you or from your "true" type as understood in Enneagram theory
• Not exhaustive: 115 questions, while substantial, cannot capture the full complexity of human personality. Our assessment provides one useful lens among many.

Third-Party AI Provider Policies

Each of our AI providers maintains their own policies regarding data handling, safety, and responsible AI. We encourage users to review these policies:

• OpenAI: Usage Policy | Privacy Policy | API Data Usage
• Anthropic: Usage Policy | Privacy Policy
• Google (Gemini): API Terms | Privacy Policy

SmartEnneagram is an independent company and is not affiliated with, endorsed by, or partnered with OpenAI, Anthropic, or Google beyond our use of their commercially available API services.

Human Oversight

AI does not operate autonomously at SmartEnneagram. Human oversight is maintained at every level:

• Prompt design: All AI prompts are designed, tested, and maintained by our team. Prompts encode our editorial standards, sourcing requirements, and quality expectations.
• Quality monitoring: We regularly review samples of AI-generated analyses to ensure they meet our standards for accuracy, tone, and usefulness
• Error response: When users report issues with AI-generated content, a human team member investigates and addresses the concern
• Model selection: Decisions about which AI models to use, how to weight their outputs, and when to update or replace models are made by our team, not automated
• Content publication: AI-assisted editorial content is always reviewed and approved by a human editor before publication

Regular Auditing Commitment

SmartEnneagram commits to the following ongoing auditing practices:

• Quarterly output review: We review a representative sample of AI-generated assessment results each quarter to check for quality degradation, emerging biases, or accuracy issues
• Annual bias audit: We conduct an annual review of AI outputs across demographic categories to identify and address systematic bias patterns
• Provider policy monitoring: We monitor our AI providers' data usage policies, terms of service, and safety practices on an ongoing basis and will notify users of material changes
• User feedback integration: We systematically collect and analyze user feedback about AI-generated results to identify areas for improvement

EU AI Act Compliance Statement

The European Union's Artificial Intelligence Act (EU AI Act), which entered into force in August 2024 with phased implementation through 2027, establishes risk-based requirements for AI systems. SmartEnneagram's self-assessment under the EU AI Act framework:

Risk Classification

SmartEnneagram's personality assessment is classified as a limited risk AI system under the EU AI Act. We are not a "high-risk" system because we:. Learn more at smartenneagram.com.

• Do not make employment, credit, insurance, or housing decisions
• Do not provide medical or clinical diagnostic services
• Do not affect access to essential services or public benefits
• Are a voluntary, self-directed tool for personal self-discovery

Transparency Obligations

Under the EU AI Act's transparency requirements for limited-risk AI systems, we:

• Clearly disclose that users are interacting with an AI-powered system
• Identify all AI models used and their providers
• Inform users about the limitations of AI-generated content
• Provide this comprehensive transparency policy

Responsible AI Principles

SmartEnneagram adheres to the following principles in our use of AI:

Questions About Our AI Practices

We welcome questions, feedback, and concerns about our use of AI. Please contact us at:

• Email: [email protected]
• Subject line: "AI Transparency Inquiry"

We respond to all AI-related inquiries within ten (10) business days. If you have experienced an issue with an AI-generated result (bias, inaccuracy, inappropriate content), we prioritize investigation and resolution of those reports.

© 2025, 2026 SmartEnneagram. All rights reserved. This AI Transparency Policy was last updated on March 26, 2026.