Context hints

In ChatGPT ads you tell OpenAI when an ad group is relevant by writing context hints: short descriptions of relevant products, use cases or customer needs. OpenAI states that context hints are not exact-match keywords, and an ad group can hold up to 2,000 of them.

The approach fits the placement. A ChatGPT ad appears below a response, not on a results page for a typed query, so the useful signal is the situation a conversation describes. A hint such as “planning a week of vegetarian meals on a budget” captures a need that no single keyword does. Qualifying advertisers can add negative phrases to exclude brand-specific placements.

Hints work alongside the other ad-group settings: countries and location IDs, and platform targeting for the iOS app, the Android app and the web. They do not change what ChatGPT says. OpenAI states that ads run on separate systems and that advertisers can’t shape responses.

Write hints the way a customer would describe the problem, not as a list of product names, and put distinct use cases in separate ad groups so reports show which situation earned the clicks. Context hints decide relevance, not approval: the ad itself must still pass OpenAI’s review before it serves. In Admux, the ChatGPT leg of a campaign carries the targeting you set in the draft, and GET /channels lists which targeting options Admux supports for each channel, so check it before you plan around a specific field. A draft that asks for targeting a channel can’t take returns a TARGETING_UNSUPPORTED error.