Advertise to AI agents.
When an AI agent acts on someone's behalf, your ad should be visible to both. Here's how.
The AI agent is the new intermediary.
AI agents increasingly mediate purchases, research, planning, coding, and support. A shopper asks an assistant to compare running shoes. A product manager asks a research agent to map a competitive landscape. A developer hands a coding agent a ticket and watches it produce a pull request. A traveler tells a planning chatbot to assemble a long weekend. In each case, the agent stands between the person with the intent and the action that satisfies it.
Brands that show up at the agent layer show up at the decision layer. The old funnel had an obvious surface to advertise against — the web page the user opened, the feed they scrolled, the results page they skimmed. The agent flattens all of that. The user asks a question, and the agent returns an answer. If your brand is not part of the answer in a disclosed, legible way, you are not in the consideration set.
There are two distinct jobs inside this idea, and treating them as one is the first mistake advertisers make. The first job is reaching the human who uses the agent. That person is still a human with a budget, a preference, and the ability to click. They see the output. They read it, watch it, listen to it, or run it. An ad in that output is an ad to a person. The second job is being discoverable to the agent itself as it performs tasks — being available as a resource the agent can cite, call, or recommend without a human reading anything in the moment.
Surfacedd addresses the first job today. Our network places disclosed sponsored Surfaces inside AI outputs that humans see. The second job is where the category is heading, and it is a different problem with different rules. The rest of this page separates the two, so your team can spend on the one that matches the behavior you want.
The two kinds of agent advertising.
If you read what advertising for AI agents means as a category, you already know the shape. This page gets concrete about what a buyer can act on today and what a buyer should plan for next. The two kinds split along one axis: who is looking at the output.
Human-visible Surfaces.
Disclosed ads that the human user sees inside the AI app. Text Surfaces next to chatbot answers. Image Surfaces inside generated scenes. Voice Surfaces before or after a spoken reply. Code Surfaces in the comment layer of a completion. This is what Surfacedd does today. Ads travel with the user’s experience of the AI product, not alongside it in a banner slot that the AI app does not have.
The unit is the Surface. The disclosure is structural — the label is part of the Surface itself, not an optional attribute the SDK may or may not render. When a Surface ships, it ships labeled. Buyers looking to reach AI users buy here. Buyers curious about the specific inventory can review the AI brand placement formats in detail.
Agent-visible discovery.
Being represented in a way agents can find and cite as they perform tasks. When a research agent compiles a comparison, it pulls from sources the model or the tool chain can reach. When a coding agent recommends a library, it pulls from the registries and documentation it has access to. Being present and well-described in those inputs is a category of visibility that did not exist in web advertising.
Surfacedd does not pay for this. Paying an agent to cite you would cross the line from placement to edit — the line our manifesto draws and keeps. A network that promises “get your brand mentioned by ChatGPT in exchange for payment” is selling manipulation, not advertising. What brands can do is register themselves as disclosed, agent-available resources through standard mechanisms: public sitemaps, structured data on their own properties, MCP servers where that pattern fits their product, and well-maintained documentation that the model providers index. This is a discovery problem, not a placement problem. It is solved by being a good source, not by buying a slot.
The practical advice is simple. Fund human-visible Surfaces as a media line item, the way you fund any other advertising channel. Fund agent-visible discovery as a brand infrastructure line item, the way you fund SEO, developer relations, or PR. They belong to different teams and they report against different metrics. Combining them into a single bucket produces bad briefs and worse attribution.
Formats and where they run.
Surfacedd runs four Surface types across a network of AI apps. Here is how each one maps to brand categories. The examples below are illustrative of typical campaigns, not an exhaustive list of the network.
DTC and consumer goods
Image and text Surfaces in generative design and lifestyle apps. A user prompts an interior scene, a kitchen layout, an outfit, a moodboard — the output includes a disclosed sponsor product in-frame, credited in a line under the image. Text Surfaces run in the recommendation and how-to apps the same audience uses to decide what to buy.
Travel
Text Surfaces in chatbot planning tools, where a user describes a trip and receives an itinerary. Image Surfaces in travel-inspiration image generators, where a hotel, an airline, or a destination brand appears in a disclosed scene. Voice Surfaces in audio guides and driving-mode assistants where a short disclosed sponsor segment fits the modality.
SaaS and developer tools
Code Surfaces in completion tools, where a disclosed sponsor line appears in the comment layer of a generated snippet or as a sidecar suggestion the developer can accept, dismiss, or ignore. Text Surfaces in technical chatbots where a user asks how to implement something and the answer is followed by a labeled sponsor recommendation.
Financial services
Text Surfaces in research and planning chatbots, subject to the disclosure rules that apply to regulated categories in your market. Creative review is required and jurisdictional gating is supported. We do not run financial Surfaces where the surrounding context is advice that should come from a licensed professional.
Healthcare and pharma
Limited. Voice Surfaces run in general-wellness assistants where the content is not clinical. We do not place healthcare or pharma Surfaces in clinical decision support tools, patient-facing symptom triage, or any context where the agent is answering a medical question. Disclosure alone is not enough in those contexts; the right answer is not to place.
What we won't do.
The short list, restated so the commitment is on the page you are reading, not only on the one it links to.
- No undisclosed placements. Every Surface ships labeled as sponsored.
- No purchased organic mentions. We will not take money to alter what the model says in its own voice.
- No private-conversation targeting. The contents of a user’s chat are not an input to ad selection.
- No political ads. Not in election windows, not between them. The category sits outside this network.
- No placements targeting minors. Surfaces do not run in apps whose audience is primarily under 18, and age-gated contexts are enforced on the supply side.
These are not policies we will reconsider for a large enough buy. They are the shape of the network. Read our disclosure principles for the full manifesto and the reasoning behind each line.
Frequently asked questions.
Can I buy a mention inside ChatGPT's organic answer?
What's the difference between an agent surface and a chatbot surface?
How do I brief a campaign for an agent audience?
Is there a difference in pricing between agent and chatbot placements?
What about MCP-enabled agents?
Reach users inside the AI tools they already use.
CPC from $0.50. CPM from $5. Text, image, voice, and code placements across independent AI apps.