Glossary
What is an AI DSP?
An AI DSP is a demand-side platform built for buying inside LLMs and AI assistants, optimising around prompt relevance, conversational context and intent strength rather than impressions, cookies and audience segments.
What an AI DSP does differently
A conventional DSP answers one question very fast: given this impression and what I know about this user, what should I bid? The inputs are an identifier, a page context and an audience segment.
An AI DSP answers a different question. There is no durable identifier, and the page context is a live conversation. What it has instead is the user's stated intent, in their own words, at the moment they stated it. Bidding is therefore against the strength and specificity of that intent rather than against the value of a known individual.
In practice that inverts the optimisation problem. A conventional DSP finds the right person and accepts whatever context they are in. An AI DSP finds the right moment and accepts whoever is in it.
What it optimises on
- Intent strength. How close the conversation is to a decision — see intent density.
- Prompt relevance. How well the advertiser's offer answers what was actually asked, not what category it belongs to.
- Conversational position. A recommendation early in exploration and one at the point of choosing behave very differently.
- Eligibility and safety. Whether the conversation is one an advertiser should appear in at all, resolved before bidding rather than after.
- Frequency across the conversation, not across pageviews — see conversational frequency cap.
The latency constraint
A display auction runs while the page is loading and has roughly 100ms. An AI DSP has to return a decision inside the window in which the assistant is generating its response — the user is already watching text appear, and a late answer is simply not rendered.
That is a harder engineering constraint than display, and it shapes the architecture. Evaluation has to be cheap enough to run inline, which favours pre-computed eligibility and compact context representations over heavyweight per-request modelling. Our ad infrastructure page covers how the decision path is arranged.
Do you need an AI DSP?
For most advertisers today, no — an AI ad network or a managed buy reaches the same inventory with far less operating overhead. A DSP earns its keep when you are running enough volume that bid-level control changes the outcome, and when you have the team to use it. Our self-serve and managed service options set out where the line usually falls.
Common questions
Can my existing DSP buy AI inventory?
Generally not natively, because the standard bid request has no way to describe a conversation and the latency budget does not fit inside response generation. Some are adding paths to it; today most AI buying runs through platforms built for the surface.
How does an AI DSP target without cookies?
It targets the conversation rather than the person. The signal is what the user has just said they are trying to do, which is both stronger than a third-party segment and unaffected by identifier deprecation.
More in ai and conversational advertising
How buying, selling and measurement work when the surface is a conversation.
- AdCP (Advertising Context Protocol)
- Agent-to-Agent Attribution (A2A attribution)
- Agentic advertising
- Agentic conversion
- AI ad network
- AI media
- AI open web
- AI publisher
- AI share of voice (AI SOV)
- AI SSP (AI supply-side platform)
- AI walled gardens
- Answer insertion
- Artificial intelligence (AI)
- Branded agent
- ChatGPT Ads
- Citation bidding
- Conversational ads
- Conversational frequency cap
- GEO / AEO (Generative engine optimization / Answer engine optimization)
- Hallucination liability
- Intent density
- Machine learning
- MCP (Model Context Protocol)
- Post-click ad economy
- Prompt inventory
- Sponsored follow-up
- Sponsored intelligence (SI)
- Sponsored prompt
- Synthetic audience
- Trust graph