What are conversational ads?

What makes an ad conversational

Two properties, and both are necessary. The placement occurs inside a dialogue rather than beside static content, and the match is made against live conversational context rather than against a page category or a stored profile.

A banner rendered in a chat window is not a conversational ad; it is a banner in an unusual location. The distinguishing feature is that the system knows what the user just asked and places on that basis.

The formats

• Sponsored card. A labelled result rendered beneath the assistant's answer, visually separated from it. The dominant format, and the one ChatGPT Ads uses.
• [Sponsored follow-up](/resources/glossary/sponsored-follow-up). A paid suggested next question that extends the conversation in a brand-relevant direction.
• [Sponsored prompt](/resources/glossary/sponsored-prompt). A paid entry point that opens a conversation on a brand-relevant topic.
• Inline recommendation. The brand named within the answer itself, disclosed — see sponsored intelligence and citation bidding.

What the creative actually is

This is the largest practical break from display. There is no artwork to traffic. An assistant renders in its own voice and layout, so what an advertiser supplies is structured content: what the product is, who it suits, what it costs, what it does not do, and where to go next.

That inverts the creative process. The work is not designing an execution; it is writing accurate, specific, machine-usable claims that survive being restated by a model in a context you did not choose. Vague positioning language has nothing to render. Our guide to what actually gets shown covers the mechanics in more detail.

The constraint that defines the format

Every conversational format is bounded by the same rule: the advertising cannot change the answer. Once a user suspects the assistant's recommendations are for sale, the product loses the trust it is monetising, which is why the separation is architectural rather than a policy.

The commercial consequences follow from that. Supply is deliberately scarce, relevance thresholds are high, frequency is capped per conversation rather than per pageview — see conversational frequency cap — and brand safety is a gate rather than a filter applied afterwards.