Frequently Asked Questions¶

General Questions¶

What is Mosaic?¶

Mosaic is a Stripe-like API infrastructure for delivering contextually relevant ads within AI. Our platform connects publishers (AI model providers) with advertisers looking to reach users through Model-as-a-Platform (MAAP) applications.

How does Mosaic work?¶

Mosaic works by matching keywords from user conversations with relevant ad campaigns. When an AI model integrates with our API, it sends conversation context and keywords to our system, which then returns contextually appropriate advertisements.

What types of AI models can integrate with Mosaic?¶

Any AI model or application that engages in text-based conversations with users can integrate with Mosaic. This includes chatbots, virtual assistants, AI-powered customer service tools, large language models (LLMs), Small language models (SLMs), and other AI-powered applications.

How do I get an API key?¶

API keys are provided to publishers when they register their AI models with Mosaic. Sing-up for a Mosaic account: HERE

Can I use the same API key for multiple models?¶

No, each AI model should have its own unique API key. This allows for accurate tracking and analytics for each model independently.

Is there a rate limit for API calls?¶

Currently, there are no strict rate limits implemented. However, we recommend implementing reasonable throttling in your application to ensure optimal performance.

What happens if no relevant ads are found?¶

If no ads match the provided keywords, the API will return a 404 error with the message "No matching advertisements found." Your application should handle this gracefully.

How should I handle errors from the API?¶

Implement proper error handling in your application to gracefully manage cases where the API returns errors. See our Error Handling documentation for details on specific error codes and messages.

Integration Questions¶

Can I customize the appearance of ads?¶

Currently, ads are delivered in a standard markdown format. While you can't customize the content of the ads, you can control how they're displayed in your application by styling the rendered markdown.

How are clicks tracked?¶

Clicks are tracked automatically when users interact with the provided campaign_link. This link includes tracking parameters that record the click event before redirecting to the advertiser's target URL.

Do I need to track impressions manually?¶

No, impressions are automatically tracked when you call the /api/fetch-ads endpoint. No additional implementation is required for basic impression tracking.

Data Collection and Privacy¶

What data do we collect from MAAP partners?¶

We collect non-PII contextual data that helps match ads to content effectively: - User queries & context (e.g., what question was asked, the response type, sentiment) - Session metadata (e.g., AI model used, conversation topic, engagement levels) - Interaction signals (e.g., how users respond to ad placements—clicks, dismissals, time spent) - Device & locale (basic geo signals but without personal tracking)

What information do we store about the user?¶

We store a contextual "Data Graph" rather than traditional user profiles with PII. The Data Graph includes: - Queries & topics of interest (e.g., "User asked about productivity tools, recommended Notion") - Engagement signals (e.g., Clicked on an ad, ignored, asked for alternatives) - Session metadata (e.g., Model used, conversation intent, previous engagements) - Sentiment analysis (Was the user enthusiastic, neutral, or negative about a suggestion?) - Aggregated behavioral insights (Do they often inquire about travel? Tech? Fashion?)

How do we target users?¶

• Real-time intent recognition: Instead of pre-tracking users, Mosaic analyzes their live AI interactions to determine ad relevance.
• Unlike traditional web tracking, we don't track users across the web but rather capture conversational intent at the moment.
• We don't scrape personal data but can enhance targeting using:
• Publicly available insights (e.g., trending topics in a given domain)
• AI model interactions (e.g., what kind of responses users engage with)
• First-party signals from MAAP partners (assuming they choose to share them with Mosaic)

How do we aggregate user data across models?¶

We use what we call the "Lily Pad Concept": - If a user interacts with multiple AI models (e.g., Anthropic's Claude and our AdSavvy model), Mosaic utilizes a cross-platform identity mechanism without violating user privacy. - Instead of tracking users directly, we assign anonymized "engagement fingerprints" based on behavior patterns. - We use hashed session identifiers that allow us to recognize recurring user behavior across different AI models without storing PII. - Cross-model consistency: We standardize how different AI platforms share engagement data securely. - User control & transparency: Users can opt-in for preference-based ad serving rather than being auto-tracked. - Privacy-first approach: No personal identities—just behavior-based anonymized intent graphs.

Business Questions¶

How is revenue shared with publishers?¶

Revenue sharing is based on a percentage of the advertising spend. The exact revenue share percentage depends on your specific agreement with Mosaic. Contact our team for details at Support

How do I view my earnings and performance metrics?¶

Publishers can access a dashboard that shows impressions, clicks, click-through rates, and earnings. Login credentials are provided when you register with Mosaic.

Can I control which types of ads appear in my application?¶

Currently, ads are matched based on user context and keywords only. Future versions of the platform will include more granular controls for publishers to filter ad categories.

How often are payouts processed?¶

Payouts are typically processed monthly, with a minimum payout threshold. Specific payout terms are outlined in your publisher agreement.

Support and Resources¶

Where can I get help with integration issues?¶

For technical support, contact our team at Support.

Is there a testing environment available?¶

Yes, we provide a sandbox environment for testing your integration before going live. Contact our team at Support.

Where can I report bugs or suggest features?¶

You can report bugs or suggest features by emailing Support.