search_experiments
Brands' own A/B tests and which version won. Use it when the user asks what brands tested, which video, text or hook won, or for 'A/B test' or 'split test' results ('did shapewear brands in France test different videos with the same text, and which won?'). A test = ads of one brand launched within 14 days that share one element and differ in another: same video with different texts, same text with different videos (most tests), or same script with a different hook (rare). Each version: its ads with links, launch day, running or stopped, days live, EU reach, and the text that differs, whether the test is clean (only the tested element differs; this compares ad fields, not the video content: right 35 of 40 times in a hand check, so say it), plus a verdict: the version that reached most (with the ratio) when the gap is 2x or more; when the only signal is that one version stopped first, it says so and that this is not a confirmed outcome yet (repeat that). Only tests with a measured outcome are served (a version still running while another stopped, or 2x+ the reach). A summary covers all matching tests per kind (decided by survival, by reach, median reach ratio), shown only from 30 tests of a kind. Costs 5 credits (0 if nothing matches).
Tool call template
This tool declares no required parameters. Add optional parameters from the input schema as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {},
"name": "search_experiments"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| kind | No | ||
| limit | No | default 10 | |
| niche | No | as in search_winners (see filters) | |
| format | No | Meta's display format of the ads tested, e.g. VIDEO, IMAGE | |
| offset | No | offset + limit at most 100: the first 100 results of a query | |
| country | No | 2-letter code: tests with an ad reaching that country | |
| with_outcome | No | always true: only tests with a measured outcome are served |