How to measure brand AI ads with holdout tests | Particular Audience
Hold back a random share of matching requests from the sponsored placement and compare sales between exposed and held-out shoppers. The difference is.
How to measure brand AI ads with holdout tests
Hold back a random share of matching requests from the sponsored placement and compare sales between exposed and held-out shoppers. The difference is the incremental effect.
Report separately
Keep AI-answer impressions, clicks and sales separate from site inventory, and report per retailer where the feeds allow it.
Run it with Particular Audience
The Brand AI Commerce Programme lets a brand build its own AI discovery app, list participating retailer feeds for live price and stock, and fund the Sponsored Prompts that surface the app. Adaptive Transformer Search and the open-source ATS MCP server match relevant products to relevant ad requests by meaning. Brands are onboarded in cohorts by category; apply here.
Sources
- Adaptive Transformer Search MCP server (GitHub), checked 2026-10-25
How Particular Audience does it
Frequently asked questions
Does a bid get a brand into an irrelevant answer?
No. Only requests that score as a strong match by meaning are eligible; the bid orders relevant ones.
Who handles checkout?
The retailer the shopper chooses.
Related guides
- What are sponsored prompts?: Sponsored prompts are a retail media format for AI conversations. When a shopper asks a retailer's assistant, voice service or AI search for something,.
- Sponsored products in agentic and conversational AI: How retailers run sponsored products inside their own AI assistant, ChatGPT app or Claude connector: relevance-first ranking, labelling, CPM and CPC pricing, pacing and holdout measurement.
- Brand discovery app: A brand discovery app is a brand-owned conversational app that helps shoppers find ideas and products from the brand, with live availability at.