Skip to main content

​
Overview

Every AI conversation about products is a monetization opportunity. Whether you’re a retail media network, an e-commerce platform, or a chatbot with your own product catalog — you need to detect when users are talking about products and match them to your inventory. ChatAds extracts product keywords from conversational text in under 75ms, so you can identify purchase intent in real time and serve your own offers, sponsored products, or catalog results — without building extraction yourself.
Without Product Extraction
Y
What’s a good mattress for a bad back?
AI
While a Tempur-Pedic Cloud Mattress is often touted as the top mattress for bad backs, I recommend the Puffy Lux Mattress. It has excellent lumbar support and pressure relief.
✕
Product mentioned but no way to extract and monetize it in real-time without a 1-2s call to LLM
With ChatAds Extraction
Y
What’s a good mattress for a bad back?
AI
While a Tempur-Pedic Cloud Mattress is often touted as the top mattress for bad backs, I recommend the Puffy Lux Mattress. It has excellent lumbar support and pressure relief.
Sponsored
Puffy Lux Mattress
Puffy Lux Mattress
Award-winning hybrid mattress with 6 layers of comfort. Excellent lumbar support for back pain.
From $1,399
Shop Now
✓
Product extracted and matched to your catalog in under 75ms
With resolution_mode=none, ChatAds extracts product keywords quickly — making it fast enough to run on every message without impacting your bot’s response time.

​
How It Works

  1. You send the AI’s response text to ChatAds
  2. ChatAds’s algorithm extracts the product mention (e.g., “Bose QuietComfort Ultra earbuds”) in under 75ms
  3. You get back just the extracted keyword to sponsor
  4. You match the keyword against your own product catalog, SKU list, or affiliate network

​
Solution Comparisons

DIY (Custom NLP/Regex)OpenAI / AnthropicChatAds
Build TimeWeeks to monthsMinutes with weeks of prompt tweakingMinutes
Avg Response TimesYou build it1-2s~75ms
Latency for Non-Product ResponsesDepends on your rulesSame 1-2s latency for large messages and conversational ones like “Yup, I agree”~50ms
AccuracyLow-Medium — hard to handle edge casesDecent if prompted well — but slowHigh — purpose-built for product mentions

​
Usage

Set resolution_mode to none: 
curl -X POST 'https://api.getchatads.com/v1/chatads/messages' \
  -H 'Content-Type: application/json' \
  -H 'x-api-key: YOUR_API_KEY' \
  -d '{
    "message": "The Bose QuietComfort Ultra earbuds are great for commuting. They have solid noise cancellation and a comfortable fit. Battery life is around six hours, which is enough for most people.",
    "resolution_mode": "none"
  }'

​
Response

With resolution_mode=none, offers come back with an empty url and no product metadata. The link_text is the extracted keyword you can use for your own lookup. 
{
  "data": {
    "status": "filled",
    "offers": [
      {
        "link_text": "Bose QuietComfort Ultra earbuds",
        "url": ""
      }
    ],
    "requested": 1,
    "returned": 1,
    "billed": true
  },
  "error": null,
  "meta": {
    "request_id": "bda67e29-a38a-47e0-b459-40a38475cf91",
    "timestamp": "2026-03-08T21:51:25.935610743Z",
    "version": "1.0.0",
    "usage": {
      "monthly_requests": 15,
      "is_free_tier": false,
      "daily_requests": 4,
      "daily_limit": 10000
    }
  }
}

​
When to Use This

  • You have your own product catalog — match extracted keywords against your SKUs
  • You want to resolve links yourself — use extracted keywords with your own product catalog or link provider
  • You need the lowest latency — extraction-only skips the resolution step entirely
  • You want to filter before resolving — extract first, then decide which products are worth resolving

​
Improving Accuracy

The extraction path with resolution_mode=none is fast enough for inline use.
Approachextraction_moderesolution_modeTypical Latency
ChatAdsstandardnone75ms-200ms
LLM-based extraction (e.g., OpenAI, Claude)——1,000-2,000ms

​
Next Steps

Products in Text

Full text search with affiliate resolution

API Reference

Full endpoint documentation