Glossary

What is an AI SSP?

An AI SSP is a supply-side platform that helps AI publishers package, price and sell their conversational inventory to advertisers and AI DSPs, managing ad eligibility, context signalling and yield.

AI and conversational advertising

All 220 terms

What an AI SSP does

A display SSP takes a publisher's slots, exposes them to demand, enforces floors and maximises yield. An AI SSP does the same job for inventory that has no slots.

The work divides into three parts. It decides which conversations are eligible to carry advertising at all. It describes an eligible moment to buyers in a way they can bid against without receiving the conversation itself. And it manages the trade-off between revenue and experience on the publisher's behalf.

Eligibility is the core function

On a web page, essentially every pageview is sellable and the SSP's job is pricing. On an AI surface, most conversations should never carry advertising, and deciding which ones may is the platform's most consequential responsibility.

The rules are part commercial and part editorial: exclude sensitive topics outright; require a threshold of commercial relevance; suppress advertising where the user is in distress, seeking medical or legal guidance, or working on something private. Getting this wrong is not a brand safety incident in the display sense — it is a failure of the product the publisher is monetising.

Describing inventory without leaking it

A bid request cannot contain the conversation. It is the user's private text, and sending it to every bidder would be indefensible.

So the AI SSP has to abstract: a topic, a measure of intent density, a stage in the decision, a set of eligibility flags. Enough for a buyer to value the opportunity, not enough to reconstruct what was said. How well a platform does this reduction largely determines what its inventory clears at, because a buyer who cannot assess relevance will bid conservatively or not at all.

Yield in a scarce market

  • Floors matter more than fill. Inventory is scarce and non-renewable; selling it cheaply to raise fill rate destroys more value than it captures.
  • Frequency is a yield lever. Capping protects the session, and a protected session produces more monetisable moments over time.
  • Demand mix over demand volume. A small set of advertisers genuinely relevant to the surface will outperform a large undifferentiated pool.
  • Measure per query, not per impression — see revenue per query.

Common questions

Do I need an AI SSP, or can I sell directly?

Direct works while you have a handful of advertisers and someone to manage them. An SSP earns its place when you want competitive demand on every eligible moment without running that auction yourself.

Does the SSP see my users' conversations?

It should not need the raw text. A well-designed integration passes derived signals — topic, intent strength, eligibility — and keeps the conversation on the publisher's side.

More in ai and conversational advertising

How buying, selling and measurement work when the surface is a conversation.