Context Hints vs. Keywords
How ChatGPT ads targeting is a genuinely different mental model from search — not just a rename of the same mechanic.
The old model: bidding on text
In search advertising, you choose keywords, set a bid, and win an auction when someone's query matches your keyword closely enough — exact, phrase, or broad match. The system is explicit: you can see exactly which term triggered your ad, and a keyword that gets no traffic is easy to diagnose.
The new model: describing meaning
ChatGPT ads don't have an auction on text. Every ad group carries one context hint — a description of the audience, intent, and situation — and OpenAI's system compares the meaning of that hint against the meaning of the live conversation. There's no bid to raise if a hint isn't triggering; the fix is always to rewrite the hint so it describes something more specific.
Side by side
| Keywords | Context hints | |
|---|---|---|
| Matching | Text (exact/phrase/broad) | Meaning |
| Fallback if too narrow | Broad match still catches some traffic | None — a vague hint just doesn't show |
| Unit of targeting | A list of keywords | One sentence per ad group |
| What to optimize | Bids, match types, negatives | Specificity of the description |
Why this changes how you write targeting
The keyword instinct is to cast a wide net and refine with negatives. That instinct works against you here — a context hint rewards going narrower, not broader, because specificity is what lets ChatGPT recognize the conversation it's meant for.
Try it on your product
The generator writes a context hint grounded in real ChatGPT ad data — free, no sign-up to try it.