AI Search with Sponsored Text Weaving - Technical Overview

⚠️ Important: This use case is about weaving sponsored text into LLM responses, not displaying separate visual product cards. For visual product ads, see AI Search Product Ads.

📺 Live Demo: See this implementation in action at https://demo-aura.affinity.net/ai-llm-weaving

Executive Summary

This document outlines how AI publishers can weave sponsored text naturally into LLM-generated responses using the AdCP (Ad Context Protocol) framework. When users search on your AI platform, your LLM (provided by you, the AI publisher) generates answers that include relevant sponsored mentions alongside organic recommendations. Users can click these mentions to see detailed information.

Key Distinction: This is about text integrated within the LLM response, not separate visual product cards. The LLM weaves sponsored brand mentions naturally into its answer text.

Key Benefits:

Natural Integration: Sponsored text woven seamlessly into LLM responses
LLM Control: Your LLM decides when and how to weave sponsored content
Expandable Details: Users can click to see full information in a modal/panel
Non-Disruptive: Appears as natural recommendations, not intrusive ads
AdCP Framework: Standardized protocol built on OpenRTB 2.6
Engagement-Based: Revenue from user engagement (expansions, clicks)

Compare With:

AI Search Product Ads displays visual product cards separately using IAB Native Ads
AI Search SEM Text Ads displays text ads in search results format

Use Case Example

User Query: "What are the best family-friendly golf resorts in California?"

AI Response with Sponsored Content:

California has several excellent family-friendly golf resorts:

Pebble Beach Resort offers world-class golf with family amenities including
kids' programs and beach access.

[Palm Springs Golf Resort] features supervised kids' activities while parents
enjoy championship courses designed by Jack Nicklaus, with all-inclusive
packages starting at $299/night.

Torrey Pines in San Diego combines stunning coastal views with excellent
junior golf programs.

User Interaction:

Sponsored mentions are subtly highlighted (e.g., gentle background color)
Clicking expands to show: full description, images, pricing, booking CTA
Non-intrusive: blends naturally with organic recommendations

Why AdCP for AI Search?

1. Standardized Integration

Built on OpenRTB 2.6 (industry standard for programmatic advertising)
No custom integrations needed for each advertiser
Works with existing ad tech ecosystem (DSPs, SSPs, ad servers)

2. Flexible Creative Assembly

Advertisers provide structured assets (text, images, URLs)
Your LLM weaves content naturally into responses
Format definitions ensure consistency and quality

3. Privacy-First Design

No PII required in bid requests
Contextual targeting based on search intent and topics
GDPR/CCPA compliant by design

4. Rich Analytics

Track impressions (when sponsored content appears)
Track expansions (when users click to see more)
Track clickthroughs (when users visit advertiser site)
Universal macros provide consistent tracking across publishers

5. Quality Control

Format specifications define asset requirements
Validation ensures quality before serving
Your LLM decides relevance and placement

Architecture Overview

Note: The LLM is provided and controlled by the AI Publisher. The AI Publisher uses their own LLM to generate responses and weave sponsored content naturally into search results.

Format Specification

AI Search Sponsored Message Format

Format ID: ai_search_sponsored_message_v1

Type: Contextual weaving with expandable details

Key Features:

Brief inline mentions (60-80 characters)
Expandable detailed view on click
Visual assets (logo, hero image)
Pricing information
Clear call-to-action

Required Assets

Asset	Purpose	Max Length
`brand_name`	Brand identifier for highlighting	50 chars
`brand_context`	Brief description for inline mention	300 chars
`key_highlights`	Top 3 features (pipe-separated)	3 items
`clickthrough_url`	Landing page URL	-
`impression_tracker`	Impression tracking URL	-
`click_tracker`	Click tracking URL	-

Optional Assets

Asset	Purpose	Max Length
`detailed_description`	Full description for expanded view	500 chars
`pricing_info`	Pricing details	100 chars
`call_to_action`	CTA button text	30 chars
`highlight_keywords`	Keywords to emphasize (pipe-separated)	5 items
`brand_logo`	Logo image (200x200px)	-
`hero_image`	Main image (600x400px)	-
`expand_tracker`	Expansion tracking URL	-

OpenRTB Integration

Note: We leverage the OpenRTB standard for forward-compatible protocol integration with the ad ecosystem. However, we do not offer impression-based bidding. The protocol is used for structured communication and asset delivery.

Bid Request Structure

Your platform sends standard OpenRTB 2.6 bid requests with AdCP extensions:

{
  "id": "search-req-001",
  "imp": [
    {
      "id": "imp-1",
      "ext": {
        "aura": {
          "adcpFormats": [
            {
              "agent_url": "https://creative.adcontextprotocol.org",
              "id": "ai_search_sponsored_message_v1"
            }
          ]
        }
      }
    }
  ],
  "site": {
    "domain": "aisearch.com",
    "cat": ["IAB19-30"],
    "keywords": "family golf resorts,california,vacation"
  },
  "ext": {
    "aura": {
      "intent": {
        "value": "information_seeking",
        "confidence": 0.92,
        "topics": ["family travel", "golf", "california resorts"]
      },
      "sentiment": {
        "value": "positive",
        "score": 0.75
      }
    }
  }
}

Key Context Signals:

Intent: What the user is trying to accomplish
Topics: Extracted from search query
Sentiment: User's emotional tone
Keywords: Relevant terms for targeting

Bid Response Structure

Advertisers respond with structured assets:

{
  "bid": {
    "id": "bid-001",
    "ext": {
      "aura": {
        "adcpFormat": {
          "formatId": {
            "agent_url": "https://creative.adcontextprotocol.org",
            "id": "ai_search_sponsored_message_v1"
          },
          "creativeManifest": {
            "promoted_offering": "Palm Springs Golf Resort",
            "assets": {
              "brand_name": {
                "content": "Palm Springs Golf Resort"
              },
              "brand_context": {
                "content": "Award-winning family golf resort with supervised kids' programs and championship Jack Nicklaus course"
              },
              "key_highlights": {
                "content": "Supervised kids' programs|Championship Jack Nicklaus course|All-inclusive from $299/night"
              },
              "detailed_description": {
                "content": "Full resort description with amenities, activities, and family features..."
              },
              "pricing_info": {
                "content": "All-inclusive packages starting at $299/night"
              },
              "call_to_action": {
                "content": "Book Your Family Getaway"
              },
              "clickthrough_url": {
                "url": "https://palmspringsgolf.com/book?campaign={MEDIA_BUY_ID}"
              },
              "impression_tracker": {
                "url": "https://track.brand.com/imp?buy={MEDIA_BUY_ID}&cb={CACHEBUSTER}"
              }
            }
          }
        }
      }
    }
  }
}

LLM Integration

Important: The LLM is part of your AI Publisher platform. You control the LLM and its behavior. The following describes how to provide sponsored content to your LLM for natural integration into search responses.

What Your LLM Receives

{
  "user_query": "What are the best family-friendly golf resorts in California?",
  "sponsored_content": {
    "brand_name": "Palm Springs Golf Resort",
    "brand_context": "Award-winning family golf resort with supervised kids' programs and championship Jack Nicklaus course",
    "key_highlights": [
      "Supervised kids' programs",
      "Championship Jack Nicklaus course",
      "All-inclusive from $299/night"
    ],
    "highlight_keywords": [
      "family-friendly",
      "championship course",
      "supervised kids' activities"
    ]
  }
}

LLM Instructions (Simplified)

Your LLM (which you provide and control) should:

Generate a comprehensive answer with multiple options
Naturally weave the sponsored brand among other legitimate options
Use the exact brand name provided
Incorporate 1-2 key highlights in the description
Keep the inline mention brief (60-80 characters)
Maintain a helpful, informative tone
Avoid words like "sponsored", "ad", or "paid" in the answer

Example LLM Output

California has several excellent family-friendly golf resorts:

**Pebble Beach Resort** offers world-class golf with family amenities.

**Palm Springs Golf Resort** features supervised kids' activities while
parents enjoy championship courses designed by Jack Nicklaus, with
all-inclusive packages starting at $299/night.

**Torrey Pines** combines stunning coastal views with junior golf programs.

User Experience Flow

1. Search Results Display

User sees AI-generated answer with multiple options
Sponsored mentions are subtly highlighted (gentle background color)
Visual treatment is non-intrusive and professional

2. Click to Expand

User clicks on highlighted sponsored mention
Modal/panel opens showing:
- Brand logo
- Hero image
- Full description
- Pricing information
- Call-to-action button

3. Tracking Events

Impression: Fired when sponsored content appears in results
Expansion: Fired when user clicks to see details
Click: Fired when user clicks CTA to visit advertiser site

Benefits for AI Publishers

Revenue Generation

Engagement-Based: Revenue tied to user engagement with sponsored content
Performance-Focused: Optimize for clicks and conversions
Premium Inventory: AI search commands premium rates for quality placements

User Experience

Non-Disruptive: Sponsored content blends naturally
Relevant: Contextual targeting ensures relevance
Valuable: Users get helpful recommendations
Transparent: Subtle highlighting indicates sponsored content

Technical Advantages

Standard Protocol: OpenRTB 2.6 compatibility
Easy Integration: Existing ad tech infrastructure works
Flexible: Your LLM controls weaving logic
Scalable: Handle millions of requests per day

Analytics & Optimization

Rich Metrics: Track impressions, expansions, clicks
A/B Testing: Test different weaving strategies
Performance Data: Optimize based on engagement
Revenue Reporting: Standard IAB metrics

Implementation Considerations

Quality Control

Relevance Threshold: Only weave highly relevant sponsored content
Frequency Capping: Limit sponsored mentions per session
Position Control: Decide where sponsored content appears
Fallback Logic: Handle cases with no relevant ads

Privacy & Compliance

No PII Required: Contextual targeting only
GDPR Compliant: User consent management
CCPA Compliant: Privacy string support
Transparent: Consider disclosure in expanded view

Performance

Low Latency: Bid requests add less than 50ms to response time
Caching: Cache format definitions and common assets
Async Tracking: Fire tracking pixels asynchronously
Graceful Degradation: Handle bid failures gracefully

Product Search Filtering

Overview

The affinity_ai_product_search extension enables publishers to filter product search results by price range, improving ad relevance for user queries. This is particularly useful when users specify budget constraints in their search queries (e.g., "show me iPhones under $900").

Extension Name: affinity_ai_product_search

Location: imp[].ext.affinity_ai_product_search (impression-level)

Applies To: Affillizz Search Product Ad adapter

When to Use Price Filtering

Use the product search extension when:

User query includes price constraints ("under $100", "$ 50 and above")
You want to filter products by budget range
Improving relevance for price-sensitive searches
Supporting price-based search refinement

Benefits:

More relevant product recommendations
Better match user intent with available products
Improved user satisfaction with search results
Higher conversion rates for price-appropriate products

Extension Structure

{
  "affinity_ai_product_search": {
    "min_price": 800.0,
    "max_price": 900.0,
    "price_currency": "EUR"
  }
}

Field Definitions

Field	Type	Required	Description
`min_price`	float	No	Minimum product price filter (must be non-negative)
`max_price`	float	No	Maximum product price filter (must be non-negative)
`price_currency`	string	No	ISO 4217 currency code (e.g., "EUR", "USD") for validation only

Validation Rules:

Both min_price and max_price are optional
Values must be non-negative (≥ 0)
If both set, min_price must be ≤ max_price
Invalid values are ignored gracefully (not rejected)
price_currency is used for validation only (not sent to API)

Usage Example

Complete OpenRTB Request with Product Search Extension:

{
  "id": "bid-request-123",
  "imp": [
    {
      "id": "imp-1",
      "native": {
        "request": "{\"native\":{\"ver\":\"1.2\",\"assets\":[{\"id\":1,\"required\":1,\"title\":{\"len\":80}}]}}"
      },
      "ext": {
        "affinity_ai_product_search": {
          "min_price": 800.0,
          "max_price": 900.0,
          "price_currency": "EUR"
        },
        "aura": {
          "adcpFormats": [
            {
              "agent_url": "https://creative.adcontextprotocol.org",
              "id": "ai_search_sponsored_message_v1"
            }
          ]
        }
      }
    }
  ],
  "site": {
    "id": "ai-search-sem-text",
    "keywords": "iphone 13"
  },
  "device": {
    "language": "de",
    "geo": { "country": "DE" }
  },
  "ext": {
    "aura": {
      "intent": {
        "value": "information_seeking",
        "confidence": 0.92,
        "topics": ["smartphones", "iphone"]
      }
    }
  }
}

Usage Scenarios

Scenario 1: Price Range Filter (Both Min and Max)

User Query: "Show me iPhones between $800 and$ 900"

{
  "imp": [
    {
      "id": "imp-1",
      "native": {
        "request": "{\"native\":{\"ver\":\"1.2\",\"assets\":[{\"id\":1,\"required\":1,\"title\":{\"len\":80}}]}}"
      },
      "ext": {
        "affinity_ai_product_search": {
          "min_price": 800.0,
          "max_price": 900.0,
          "price_currency": "USD"
        }
      }
    }
  ],
  "site": {
    "keywords": "iphone"
  }
}

Result: Returns products priced between $800 and$ 900

Scenario 2: Minimum Price Only

User Query: "Show me premium headphones $50 and above"

{
  "imp": [
    {
      "id": "imp-1",
      "native": {
        "request": "{\"native\":{\"ver\":\"1.2\",\"assets\":[{\"id\":1,\"required\":1,\"title\":{\"len\":80}}]}}"
      },
      "ext": {
        "affinity_ai_product_search": {
          "min_price": 50.0,
          "price_currency": "USD"
        }
      }
    }
  ],
  "site": {
    "keywords": "headphones"
  }
}

Result: Returns products priced $50 and above

Scenario 3: Maximum Price Only

User Query: "Show me budget laptops under $500"

{
  "imp": [
    {
      "id": "imp-1",
      "native": { "request": "..." },
      "ext": {
        "affinity_ai_product_search": {
          "max_price": 500.0,
          "price_currency": "USD"
        }
      }
    }
  ],
  "site": {
    "keywords": "laptop"
  }
}

Result: Returns products priced under $500

Scenario 4: No Filter (Default Behavior)

User Query: "Show me smartphones"

{
  "imp": [
    {
      "id": "imp-1",
      "native": {
        "request": "{\"native\":{\"ver\":\"1.2\",\"assets\":[{\"id\":1,\"required\":1,\"title\":{\"len\":80}}]}}"
      }
    }
  ],
  "site": {
    "keywords": "smartphone"
  }
}

Result: Returns products without price filtering (all price ranges)

Best Practices

Extension Usage

Optional: Extension is completely optional and backward compatible
Graceful: Invalid values are ignored, not rejected
Validation: Currency is for validation only (not sent to API)
Flexible: Both min and max are optional (use either or both)

Integration

Compatible: Works with existing Affinity AI extensions (AdCP, context enhancement)
Combinable: Can be combined with other impression-level extensions
Per-Impression: Applies per-impression (different filters per ad slot)
Adapter-Specific: Only affects Affillizz Search Product Ad adapter

Query Parsing

When parsing user queries for price constraints:

Extract Price Values: Identify numeric price values in query
Detect Range Type: Determine if min, max, or both
Identify Currency: Extract currency from query or use default
Validate Values: Ensure min ≤ max if both present
Add Extension: Include in bid request if price filter detected

Error Handling

Missing Extension: Request works normally without extension
Invalid Values: Ignored gracefully (e.g., negative prices, min > max)
Unknown Currency: Logged but not rejected
Partial Data: Min or max alone is valid

Integration Notes

Works With Existing Extensions

The product search extension integrates seamlessly with other Affinity AI extensions:

{
  "imp": [
    {
      "id": "imp-1",
      "native": {
        "request": "{\"native\":{\"ver\":\"1.2\",\"assets\":[{\"id\":1,\"required\":1,\"title\":{\"len\":80}}]}}"
      },
      "ext": {
        "affinity_ai_product_search": {
          "min_price": 800.0,
          "max_price": 900.0,
          "price_currency": "EUR"
        },
        "aura": {
          "adcpFormats": [
            {
              "agent_url": "https://creative.adcontextprotocol.org",
              "id": "ai_search_sponsored_message_v1"
            }
          ]
        }
      }
    }
  ],
  "ext": {
    "aura": {
      "intent": {
        "value": "purchase",
        "confidence": 0.85
      },
      "sentiment": {
        "value": "positive",
        "score": 0.72
      }
    }
  }
}

Per-Impression Filtering

Different ad slots can have different price filters:

{
  "imp": [
    {
      "id": "imp-1",
      "native": {
        "request": "{\"native\":{\"ver\":\"1.2\",\"assets\":[{\"id\":1,\"required\":1,\"title\":{\"len\":80}}]}}"
      },
      "ext": {
        "affinity_ai_product_search": {
          "max_price": 100.0
        }
      }
    },
    {
      "id": "imp-2",
      "native": {
        "request": "{\"native\":{\"ver\":\"1.2\",\"assets\":[{\"id\":1,\"required\":1,\"title\":{\"len\":80}}]}}"
      },
      "ext": {
        "affinity_ai_product_search": {
          "min_price": 500.0
        }
      }
    }
  ]
}

ADCP Protocol Components

Format Support - ADCP format framework
Creative Manifests - Manifest structure
Tracking - Tracking implementation
Contextual Weaving Format - Alternative weaving format

Context Enhancement

Context Enhancement Overview - Rich context signals
Intent Analysis - Intent detection
Sentiment Analysis - Sentiment scoring

External Standards

OpenRTB 2.6 Specification - IAB Tech Lab
IAB Content Taxonomy - IAB Tech Lab

Version History

v1.0 (2025-11-05): Initial use case guide
- AI search sponsored message format specification
- LLM weaving instructions
- User experience flow (inline → expandable)
- Implementation phases and best practices

Conclusion

AdCP enables AI publishers to monetize search results through natural, contextual sponsored messages. The standardized approach ensures:

Easy Integration: Standard OpenRTB protocol
Quality Experience: Non-disruptive, relevant content
Strong Revenue: Premium CPMs for AI search inventory
Full Control: Your LLM decides when and how to weave

By adopting AdCP for AI search, you can generate significant revenue while maintaining the high-quality user experience your platform is known for.

Executive Summary​

Use Case Example​

Why AdCP for AI Search?​

1. Standardized Integration​

2. Flexible Creative Assembly​

3. Privacy-First Design​

4. Rich Analytics​

5. Quality Control​

Architecture Overview​

Format Specification​

AI Search Sponsored Message Format​

Required Assets​

Optional Assets​

OpenRTB Integration​

Bid Request Structure​

Bid Response Structure​

LLM Integration​

What Your LLM Receives​

LLM Instructions (Simplified)​

Example LLM Output​

User Experience Flow​

1. Search Results Display​

2. Click to Expand​

3. Tracking Events​

Benefits for AI Publishers​

Revenue Generation​

User Experience​

Technical Advantages​

Analytics & Optimization​

Implementation Considerations​

Quality Control​

Privacy & Compliance​

Performance​

Product Search Filtering​

Overview​

When to Use Price Filtering​

Extension Structure​

Field Definitions​

Usage Example​

Usage Scenarios​

Scenario 1: Price Range Filter (Both Min and Max)​

Scenario 2: Minimum Price Only​

Scenario 3: Maximum Price Only​

Scenario 4: No Filter (Default Behavior)​

Best Practices​

Extension Usage​

Integration​

Query Parsing​

Error Handling​

Integration Notes​

Works With Existing Extensions​

Per-Impression Filtering​

See Also​

Related Use Cases​

Related Documentation​

ADCP Protocol Components​

Context Enhancement​

External Standards​

Version History​

Conclusion​