Nexbid for Media · Nexbid for Marketing
The agentic ad server for the DACH region
Nexbid brings content, products and intent together where people search today: on websites and in AI answers. For networks, publishers and advertisers, next to your existing ad server or in its place.
What an agentic ad server does differently
A classic ad server decides on cookies, device IDs and placements. An agentic ad server reads the context: the content of a page or the question an AI assistant is working on for someone. The result is a recommendation that fits the situation and is labelled as sponsored. Nexbid works without cookies and without fingerprinting.
Classic ad server and agentic ad server compared
Both decide which ad appears where. They differ in what they base the decision on, where they deliver and how transparent the decision is.
| Classic ad server | Agentic ad server (Nexbid) | |
|---|---|---|
| Basis for the decision | Cookies, device IDs, user profiles, placement | Context: the content of the page or the question put to the AI assistant |
| Where it delivers | Ad slots on websites and in apps | Ad slots on websites and labelled recommendations in AI answers |
| Who asks | A browser | A browser or an AI agent via the Model Context Protocol |
| Winner rule | Usually not disclosed | Published formula of bid, similarity, quality and context |
| Privacy | Consent for tracking required | No cookies, device IDs or personal profiles |
| Buying | Through interfaces and insertion orders | Through interfaces, APIs and directly by agents (MCP) |
Two operating modes
You decide how Nexbid works, per inventory if you like. If you want to keep your ad server, you keep it: Nexbid integrates. Both modes run on the same core, and a step-by-step switch with both running in parallel is supported.
Next to your ad server
Your ad server stays. Nexbid supplies it with contextual signals and agreed deals.
- Targeting keys for Google Ad Manager
- The deal as the only link between planning and delivery
- No change to your existing workflows
- Additionally monetise agentic and AI traffic, for example through your own MCP server
As your ad server
For networks that want to replace their ad server, and for publishers or advertisers without one. Nexbid delivers itself.
- Delivery through the Nexbid SDK and the discovery API
- The same contextual targeting as when running next to your ad server
- Advertiser feeds with their own shopping access for AI assistants
One targeting for every channel
The contextual targeting describes a page or question by topic, category, occasion, season and further attributes. The vocabulary is the same for every channel, the translation differs per channel.
- In use
Display
Contextual signals as targeting keys in Google Ad Manager, based on page content and structured data.
- In use
AI answers
Labelled recommendations in AI assistants via the Model Context Protocol. Buying ChatGPT ads through the official interface runs as a pilot.
- In development
Social
Social networks do not accept custom contextual signals. We analyse the briefing and set the targeting automatically in the platforms: interests, audiences and creative variants.
One platform from planning to measurement
Planning, delivery and measurement usually live in separate tools. At Nexbid they are one chain, and measurement feeds back into the next plan.
- Pilot
Briefing and planning
Analyse the briefing, derive targeting and set it up in the platforms, starting with ChatGPT ads.
- In use
Delivery
Display and AI answers with the same contextual targeting; buying in ad platforms in pilot.
- In use
Marketing mix modelling
Estimate the effect per channel with Google Meridian as the basis for budget allocation.
- In use
AI visibility
Measure whether a brand appears in ChatGPT, Claude, Gemini and Perplexity and which sources are cited.
- Planned
Touchpoint and audience analysis
Analyse the customer journey and audiences across all channels.
Next to your ad server or in its place
Most sales houses and publishers in Germany, Austria and Switzerland work with Google Ad Manager or a comparable ad server. Switching is expensive and risky. That is why Nexbid almost always starts alongside it: a tag on the page reads the context and hands it to the existing ad server as targeting keys. Campaigns, line items and billing stay where they are. What is new is that campaigns can target topics, occasions and categories without a single cookie.
If you want to replace your ad server, Nexbid can deliver itself. It has its own object model: network, advertiser, booking, placement and creative, five priority tiers from guaranteed bookings down to house ads, and the pricing models CPM, CPC and CPL. CPM bookings are billed on the signed impression beacon, not when the ad is requested. This mode is built and running as a pilot with one sales house.
Both modes can be mixed per inventory. A step-by-step switch with both running in parallel is supported: contextual signals first, then individual placements, then whole inventory areas.
- Next to Google Ad Manager
- In use. Contextual signals as key-value pairs that your line items target.
- As your ad server
- Built, in pilot. Own delivery, priorities, CPM/CPC/CPL, frequency capping without cookies.
- DSP, SSP, OpenRTB, DV360 deals
- Next step. Today the connection to the programmatic market runs through Google Ad Manager.
Monetise agentic traffic
AI assistants read pages, but they do not see banners. A recipe, a review or a guide is summarised in an answer, and the publisher earns nothing from it. That gap grows with every person who asks an assistant instead of a search engine.
Nexbid opens a second channel for this: agents ask for content and products through the Model Context Protocol. If an active campaign fits the context, Nexbid returns a recommendation labelled as sponsored next to the editorial results. The editorial order stays unchanged. If the recommendation comes about in the context of a publisher’s content, the publisher receives 90 percent of the revenue and Nexbid 10 percent.
Sponsored results only go to registered agents, meaning agents with their own identifier and a valid API key. That is access control, not cryptographic proof that a particular language model is at the other end. In addition there are Enriched Snippets: enriched product and content data that an agent pays for per call.
Contextual targeting without cookies
The targeting describes a page or a question, not a person. Nexbid reads the page content and its structured data, for example the schema.org markup of a recipe, and derives attributes from it: topic, subtopic, category, occasion, season, diet, difficulty, content type. For an agent request, the question itself is added.
There are no cookies, no device IDs and no personal profiles. The publisher SDK briefly caches contextual signals locally so the same page is not analysed again on every view; that is a cache for the content, not an identifier for the person. Frequency capping in Nexbid’s own ad server works without cookies.
The vocabulary is the same for every channel. In Google Ad Manager it becomes key-value pairs, in AI answers it is matched against the question, and on ad platforms such as ChatGPT Ads it becomes context hints for the campaign.
MCP server: your agent, your knowledge, our infrastructure
The Nexbid MCP server at mcp.nexbid.dev offers 19 tools. Search, product details, categories and the ad inventory work without a key. With an API key you add the transactions: create media buys, submit creatives, activate, pause, cancel, fetch reports, settle and trigger purchases. For agents that talk to each other rather than through tools, there is also an A2A endpoint.
You can use any MCP-capable agent: Claude, ChatGPT or your own. Your own knowledge, such as a RAG index with product data, briefings or brand guidelines, stays with your agent. Nexbid is then one tool among several that your agent calls when needed. If you do not want to run your own infrastructure, use ours: the hosted MCP server and verified brand facts as the basis for copy and recommendations.
- Your own agent
- Any MCP client calls the tools. Your RAG and your logic stay with you.
- Nexbid infrastructure
- Hosted MCP server, brand facts as a verified source, billing included.
- Open descriptions
- mcp.json, agent-card.json and adagents.json under /.well-known/ for automatic discovery.
Brand safety
When agents make recommendations, a brand needs to know which content it appears next to, and a publisher needs to know what it is serving. Nexbid addresses this in several places.
- Rules per publisher
- Allowed and blocked categories and brands, evaluated before every selection.
- Quality check on import
- Product and content data are checked as they come in, so no auto-generated nonsense ends up in recommendations.
- Labelling
- Every paid placement is labelled as sponsored and does not change the editorial results.
- Proven rule
- That a blocked ad never wins is machine-checked in Lean 4 for the model of the decision rule; the proof runs again on every change.
How your brand appears in AI answers
Whether ChatGPT, Claude, Gemini or Perplexity mention a brand can be measured. Every week Nexbid asks all four models the same questions and analyses whether the brand appears, in which position, which sources are cited and which competitors are named instead. The result is a time series rather than a hunch.
The second part is the brand space: brand facts, products, reviews and editorial content as a verified, structured source that agents access. What an agent reads there has been approved by the brand, not stitched together from a random forum post.
Channels, measurement and optimisation
An agentic ad server does not replace the other channels, which is why measurement works across channels. Marketing mix modelling with Google Meridian estimates the effect per channel and serves as the basis for budget allocation. Meta campaigns are read daily, and spend and results flow into the same report. Nexbid books ChatGPT ads through the official interface; this connection runs as a pilot.
Google Ads, LinkedIn and Display & Video 360 will be connected through the MMM project as next steps. Optimisation works on Nexbid’s own inventory today: pacing adjusts delivery to the budget every 30 minutes. Budget steering across all channels follows once there is enough campaign volume to make it reliable.
How to get started
Getting started does not require a migration. It begins with one inventory area and grows with the results. The right order depends on whether you bring content, products or both.
For publishers and sales houses, the first step is almost always running alongside Google Ad Manager: a tag supplies contextual signals, your line items target them, and within a few weeks you see which topics and occasions create demand. For advertisers it starts with the product feed and a campaign with budget and bid; from then on your products appear where the context fits, on websites and in AI answers.
- 1. First conversation
- We clarify inventory, products and goals and show which parts already run today.
- 2. Activate contextual signals
- Add the tag or connect the feed. For Google Ad Manager, the keys your line items target are created.
- 3. Open the agent channel
- Make content and products available to agents through the MCP server, with sponsored recommendations labelled.
- 4. Measure and expand
- Measure AI visibility and channel effect, then decide whether individual placements run directly through Nexbid.
The two sides of the marketplace
| Side | Brings | Gets |
|---|---|---|
| Publishers | Pages and content: articles, recipes, reviews, guides | 90% of revenue, visibility for the intent behind every visit, a connection to AI assistants |
| Advertisers | Products and an active campaign with budget and bid | Placement in the right context, paying only for actual delivery |
| Nexbid | Intent detection, the auction and open interfaces | 10% platform fee |
The Match-Moment
Advertising happens where three things meet: a piece of content, a matching product and a person’s intent. Nexbid detects that moment and serves a labelled recommendation right there.
Without content, product or intent there is no match and no revenue. A publisher alone is found but earns nothing. An advertiser alone has a catalogue but reaches no one.
An auction you can audit
The winner is decided by a published formula, not an opaque model. The rule that a blocked ad never wins is machine-checked with Lean 4, and the proof runs again on every change.
Score = 0.3 · bid + 0.3 · similarity + 0.2 · quality + 0.2 · context
Which protocols Nexbid speaks
Agentic advertising comes with many acronyms. They describe different layers and do not exclude each other.
| Protocol | What for | Who is behind it | At Nexbid |
|---|---|---|---|
| MCP (Model Context Protocol) | Connection between an AI model and tools | Open specification, initiated by Anthropic | MCP server with 19 tools |
| AdCP (Ad Context Protocol) | Vocabulary for agents buying and selling advertising, built on MCP | AgenticAdvertising.org | Nexbid is a member and implements AdCP; adagents.json is published |
| AAMP (IAB Tech Lab) | Standards for agentic media management | IAB Tech Lab | Followed; overlaps with AdCP |
| A2A (Agent2Agent) | Communication between agents | Open specification | A2A endpoint available |
| ACP / UCP | Cart and payment in AI assistants | OpenAI and Stripe / Google | Different layer: Nexbid decides what is recommended, not how it is paid for |
AdCP is not a Nexbid protocol. It is run by AgenticAdvertising.org; Nexbid is a member and implements it.
What runs today and what comes next
We separate clearly what is in use, what is built and running as a pilot, and the next steps.
- In use
Contextual signals for Google Ad Manager
Targeting keys from page content and structured data.
- In use
MCP server and labelled recommendations
19 tools, public search, transactions with a key.
- In use
AI visibility and MMM
Weekly measurement in four models, Google Meridian, Meta data in the report.
- Pilot
Own delivery as ad server
Priorities, CPM/CPC/CPL, billing on the impression beacon.
- Pilot
ChatGPT ads through the API
Create campaigns from orders, with context hints.
- Next step
Google Ads, LinkedIn, DV360 in the MMM
Connection through the MMM project for a cross-channel report.
- Next step
OpenRTB, Prebid and deals
Direct programmatic connection in addition to Google Ad Manager.
- Next step
Cross-channel optimisation
Budget steering across all channels once the volume supports it.
Who it is for
Networks and sales houses
Open up new ad space in AI answers, with or without switching ad servers.
More for sales houses →Publishers
Make your content available to AI assistants and earn from every match.
More for publishers →Advertisers and brands
Show products where intent arises, and pay only for delivery.
More for advertisers →Read more
- MCP in advertising: how an agent books through an ad server
Tools, the flow of a media buy, discovery via /.well-known.
- When the agent recommends: what MCP changes for brands
Context instead of cookies, labelling, measuring AI visibility.
- Monetising agentic traffic
For publishers and sales houses: next to Google Ad Manager, 90/10, brand safety.
- Your agent, your knowledge
Using Nexbid from Claude, ChatGPT or your own agent.
- AdCP, AAMP, MCP, ACP: which protocol does what
A guide for the DACH region.
- Classic advertising, optionally agentic
Display, video, native with or without an agent. The former Digital Campaign Manager is now part of Nexbid.
Swiss provider, operated in Frankfurt
Nexbid is a product of digital opua GmbH in Walchwil, Switzerland. The application and database run in Frankfurt. The protocol is open and published under the MIT licence.
Book a conversation
We show you how Nexbid works next to your ad server or in its place, and which parts already run today.
Send an emailFrequently Asked Questions
What is an agentic ad server?
An ad server that decides not on cookies or device IDs but on context: the content of a page or the question an AI assistant is working on. It serves labelled recommendations on websites and in AI answers.
Do I have to replace my existing ad server?
No. Nexbid can run next to your ad server and supply it with contextual signals and deals. If you want to switch, you can do it step by step and run both modes in parallel for a while.
Which channels does the targeting work in?
Today on websites through Google Ad Manager and in AI answers through the Model Context Protocol, as well as when buying ChatGPT ads. For social networks we set the targeting from the briefing automatically in the platforms; this part is in development.
How is the advertising labelled?
Every paid placement is labelled as sponsored and does not change the order of the editorial results.
Where is the data processed?
The application and database run in Frankfurt. Nexbid works without cookies and without fingerprinting; targeting relies on context, not on personal profiles.
How do publishers earn money from AI agents?
Agents ask for content through the Model Context Protocol. If a campaign fits the context, Nexbid returns a labelled recommendation with it. If it comes about in the context of your content, you receive 90 percent of the revenue.
Is Nexbid a DSP or an SSP?
Not in the OpenRTB sense today. Nexbid decides contextually, runs its own auction and connects to the programmatic market through Google Ad Manager. OpenRTB, Prebid and deals with Display & Video 360 are the next steps.
What is AdCP, and does it belong to Nexbid?
AdCP stands for Ad Context Protocol, a vocabulary that lets agents buy and sell advertising. It is run by AgenticAdvertising.org. Nexbid is a member, implements AdCP and publishes an adagents.json, but is not the author of the protocol.
Can I use my own AI agent and my own knowledge?
Yes. Any MCP-capable agent can call the Nexbid tools. Your knowledge, such as a RAG index, stays with your agent. Alternatively you use Nexbid’s hosted MCP server and verified brand facts.
How is brand safety ensured?
With rules per publisher for allowed and blocked categories and brands, a quality check on import, labelling of every paid placement, and a rule proven in Lean 4 that a blocked ad never wins.
How do I see whether my brand appears in ChatGPT or Claude?
Every week Nexbid asks ChatGPT, Claude, Gemini and Perplexity a fixed set of questions and measures whether your brand is named, which sources are cited and which competitors appear.
Which advertising channels does Nexbid connect?
Today websites through Google Ad Manager, AI answers through the Model Context Protocol and ChatGPT ads as a pilot. Meta data flows into the report daily. Google Ads, LinkedIn and Display & Video 360 follow through the MMM project.
Does Nexbid optimise campaigns in real time?
On Nexbid’s own inventory, pacing adjusts delivery to the budget every 30 minutes. Cross-channel optimisation follows once there is enough campaign volume.
How is an agentic ad server billed?
Advertisers pay per thousand impressions, per click or per lead. In Nexbid’s own ad server, CPM bookings are billed on the signed impression beacon, so only once the ad was actually shown.