What is an AI publisher?

What makes a publisher an AI publisher

The defining property is that the publisher owns the conversation, not a page. A traditional publisher sells attention against content it produced. An AI publisher sells relevance against a question the user just asked, which means the inventory is created by the user rather than by an editorial calendar.

That covers a wider set of businesses than the household names. Any application where users bring a real problem to a model — a research tool, a coding assistant, a tutor, a shopping agent, a travel planner — is an AI publisher with genuinely valuable inventory, usually far more specific than a general assistant's.

What an AI publisher controls

• Eligibility. Which conversations may carry advertising at all — the single most consequential decision, and the one that protects the product.
• [Prompt inventory](/resources/glossary/prompt-inventory). How the eligible set is described, segmented and priced.
• Format. How a paid result renders in the assistant's own voice and layout.
• Frequency. How often any one user encounters advertising across a session.
• Floor pricing and demand mix, usually through an AI SSP.

The monetisation trade-off

Every AI publisher faces the same tension in a sharper form than a web publisher did. A page can carry a bad ad and remain a good page. An assistant that gives a compromised recommendation has damaged the only thing users came for.

The practical consequence is that yield is maximised by restraint rather than by density. Publishers who have made this work run low advertising ratios, high relevance thresholds, and hard separation between the answer and the paid result. Our guidance for app builders covers the mechanics: integrating ads into an LLM chat interface and monetizing free-tier users without killing retention.

How AI publishers earn

Revenue is usually expressed per query rather than per thousand impressions, because a query is the unit the publisher actually produces and only some fraction of queries are monetisable. Revenue per query and RPM benchmarks are the two numbers most operators track, alongside whatever subscription revenue the product also carries.