ADWEEK reports that advertisers running campaigns in ChatGPT want three capabilities tightened before they scale spend: brand safety controls, prompt-matching precision, and incrementality measurement. The list, surfaced this week, mirrors complaints ad buyers lodge against every emerging format — from native to connected TV to retail media.

The headline news is the list. What each term means inside a conversational AI placement is the context.

What does "brand safety" look like in a ChatGPT reply?

In programmatic display and search, brand safety runs on a mature stack: pre-bid filtering, category exclusions, publisher block lists, and third-party verification from vendors such as DoubleVerify, IAS, and Human. None of them, as far as public reporting shows, ship native coverage for ad slots embedded in ChatGPT replies.

The surrounding text is generated at reply time. Buyers cannot pre-screen the context an ad will appear next to using a static keyword block list. Manual creative review — fine for a $50,000 pilot, unworkable for a $5 million always-on campaign — is the work-around in use today.

What does "prompt-matching" mean?

Search ads match queries to keywords. Display ads match audience segments to publisher inventory. Prompt-matching asks the system to read a user's prompt plus the model's synthesized reply and place an ad against the inferred intent.

Per ADWEEK's reporting, buyers want intent segments exposed: commerce, research, support, comparison. A user asking "best coffee maker under $200" is a different bid from a user asking how drip brewing works. Coarse topic blocks, the current fallback, flatten those distinctions and force optimization onto guesswork.

Why is incrementality the harder ask?

Incrementality measurement answers one question: did this ad drive an outcome that would not have happened anyway? Standard methods — holdout tests, geo-experiments, ghost-bid frameworks — are routine reporting on Google, Meta, and The Trade Desk.

Inside a conversational placement, each impression is a generated token inside a model reply. Buyers want measurement endpoints that connect a session to a downstream conversion without leaning on a persistent user ID. That is a tall order under European privacy rules and parts of North America, but it is the credibility threshold for moving from test budgets to scale budgets.

What remains opaque about the rollout?

OpenAI has not, to public knowledge, published a ChatGPT ad roadmap with ship dates for any of the three asks, pricing tiers beyond the closed beta, or measurement partners. That opacity is the backdrop for the complaints ADWEEK surfaced.

Holding companies handle ChatGPT as incremental inventory today, not as a search or social replacement. Without the three capabilities above — or a credible public roadmap — that posture holds.

The next open question is whether OpenAI ships the tooling, or whether competitors in the conversational AI ad space close the gap first.