Trust answer
wavebird runs the sponsor path beside your model path. Your model call stays untouched, and the placement renders in a separate slot your app controls.
wavebird is ad infrastructure for AI surfaces. It connects GenAI apps, agents, and AI surfaces to the ad market through one API, so they can fund usage without making advertising part of the model's answer. Ads render as clearly labeled placements beside the answer, never inside it. EU-hosted, data-minimizing, pilot-ready.
wavebird does not generate ad creative, copy, images, or video with AI. It does not create or manage ad campaigns, and it is not an AI ad generator. It does not insert sponsor content into prompts, context, or the model's answer. It is not an SSP, DSP, ad network, or model provider.
01
Your backend asks wavebird for a placement while the model is generating.
02
The signal is reduced to policy-safe delivery metadata.
03
Only configured partner paths can be reached.
04
Render and beacon events create the audit trail for settlement.
When your user sends a prompt, two things happen at the same time. The model thinks. An ad is matched. Neither waits for the other.
Model path
Your API call runs exactly as usual
~800ms+
Ad path
Filtered signal goes to the approved partner path
< 20ms
Filter
Only topic and language reach the ad path. No prompts, PII, or history are sent to advertising partners.
Match
Scoped partner request during model inference. No extra user-side roundtrip.
Prove
Every impression gets signed proof. Auditable. Billable. No self-reporting.
Brand safety and brand suitability
The ad path has two gates. First, wavebird decides whether the context is eligible for ads at all. Only then does it build a suitable ad-market category for matching.
Stage 1
Brand Safety
wavebird checks the context against sensitivity classes before matching. If policy marks the context as unsafe for ads, the slot becomes No-Fill and the partner path is not called.
Stage 2
Brand Suitability
Only eligible contexts continue to matching. wavebird classifies the allowed signal into an IAB content category, applies category controls, and sends only the reduced slot signal to configured partner paths.
SDK-controlled input
The SDK gives your app control over the input. Prefer a reduced topic when you have one. If prompt processing is enabled, safety and IAB matching run in the wavebird API path before partner egress.
API-controlled egress
The API path stores the raw prompt only inside the firewall boundary, erases it, and sends a reduced slot signal. Prompts, PII, and chat history do not go to ad partners.
What reaches the ad path
Your app
After firewall
topic: travel, lang: en
Ad market
Receives only abstract category and language. It never sees the original prompt, user data, or conversation history.
You control the ad path. Config is explicit. Missing config means blocked by default.
The timeline
The ad is ready before the model finishes. Zero added latency to the user.
Source: Engineering evidence documents the latency basis for the parallel ad path. Path: /resources/evidence.Proof chain
1
Slot created
2
Partner fills
3
Beacon sent
4
Proof signed
Every impression is cryptographically linked to a compute unit. Settlement is signed, auditable, and billable. No self-reported counting.
In code
server-api-flow.ts
typescript
Compliance
Built on Compute Sponsoring v1.0
Profile: CS-S (S1/P0)
The real choice is not whether ads exist. It is whether the ad path fits the GenAI compute moment, stays auditable, and can ship without rebuilding market infrastructure.
What you get
Why wavebird is different
What you get
Why wavebird sits in the middle
What you get
Why teams adopt it
Next step
Open the API docs, download the LLM integration markdown, then create a workspace to manage your publishable and secret keys. Assisted rollout stays available for enterprise or operator-led deployment.