A standardized API that sends data about brand mentions in AI search results directly from external systems without requiring a tracking script on the AI interface.
Readers interested in advanced digital analytics and brand monitoring for AI search platforms, often comparing it to other server-side collection methods.
01What it is and how it works
The Measurement Protocol works by accepting HTTP POST requests to a dedicated endpoint. Each request must include a payload with required fields such as the brand name, the source AI platform (e.g., ChatGPT, Gemini), the timestamp of the appearance, and a unique event ID to prevent duplicates. Optionally, you can include context like the query that triggered the mention or the snippet text. The platform validates the payload, checks authentication via an API key, and processes the event to update brand appearance metrics. This is similar to how Google Analytics' Measurement Protocol allows server-side data collection, but tailored for brand tracking in AI search. The protocol supports both real-time and batch sending, with a maximum payload size of 16KB per request.
It is a way to send brand appearance data from your servers to our platform using HTTP requests, so you can track mentions that are not captured by browser tags.
02What to do about it
To use the Measurement Protocol, first generate an API key from your account settings. Then, set up a server-side script that detects brand appearances in AI search results. This could be a custom crawler, an integration with a monitoring tool, or a webhook from an AI platform. When a brand mention is detected, your script sends a POST request to the Measurement Protocol endpoint with the required fields. Start with a few test events to verify the data appears in your dashboard. Monitor the response codes: 200 means success, 400 means invalid payload, 401 means authentication error. Once confirmed, you can scale up to send all relevant events. Also, ensure you handle rate limits (e.g., 1000 events per minute) and implement retry logic for failed requests.
03How it is measured or noticed
You can measure the success of the Measurement Protocol by checking the Events report in your dashboard. It shows the number of events received via the protocol, broken down by source and brand. You can also view the protocol's own logs in the API Logs section, which list each request, its status, and any errors. A sudden drop in events may indicate a configuration issue or a change in the AI platform's accessibility. Additionally, you can compare the data from the Measurement Protocol with data from other sources (e.g., browser tags) to ensure consistency. If the numbers diverge, investigate the cause.
04Common mistakes
- Sending duplicate events because the unique event ID is not generated correctly.
- Using an incorrect timestamp format (must be ISO 8601).
- Omitting required fields like brand name or source platform.
- Not handling authentication errors and assuming all requests succeed.
- Sending too many events in a short burst and hitting rate limits without a backoff strategy.
05Limits
The Measurement Protocol is not a replacement for browser-based tracking when the AI search interface is web-based and you can inject a script. It is designed for server-side or offline data where a script cannot run. It also does not apply if the AI platform does not expose brand mentions in a way your system can detect programmatically. Additionally, the protocol is often confused with webhooks; webhooks are push notifications from a platform, while the Measurement Protocol is a pull-like method where you send data. Another limit: the protocol only accepts data you send; it cannot retroactively fill gaps. You must have a detection mechanism in place.
06Worked example
Example POST request:
POST /v1/measurement
Host: api.example.com
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json
{
"brand": "Acme Corp",
"source": "ChatGPT",
"timestamp": "2025-03-21T10:30:00Z",
"event_id": "abc123",
"query": "best CRM software",
"snippet": "Acme Corp offers a comprehensive CRM solution..."
}
Response (200):
{
"status": "success",
"event_id": "abc123"
}
Frequently asked questions
Is the Measurement Protocol the same as Google's Measurement Protocol for web analytics?
No, this Measurement Protocol is specifically for sending brand mention data from AI search results, not for web analytics. It is a dedicated API for our product.
Should I use the Measurement Protocol or the browser-based tracking script?
It depends on your access to the AI search interface. Use the browser script if you can inject JavaScript; use the Measurement Protocol if you cannot, or if you want to send data from your own backend systems.
How do I send data to the Measurement Protocol?
You send HTTP POST requests to a dedicated endpoint with an API key from your account settings. The request body contains the brand mention details.
Does the Measurement Protocol work for all AI search engines?
It works for any AI search interface that you can identify brand mentions in, as long as you can send the data programmatically. It is not limited to specific engines.
What happens if I send incorrect or malformed data to the Measurement Protocol?
The API will return an error response, and the data will not be recorded. You should validate your payloads against the API documentation to avoid data loss.
How long does it take for data sent via the Measurement Protocol to appear in my dashboard?
Data typically appears within a few minutes in the Events report. However, real-time processing may have slight delays depending on system load.
Asked out loud
spoken, not typedThe same term in the words somebody uses speaking to an assistant rather than typing into a box — written from the situation, which is why each one carries the situation it came from.
Yes, you can use the Measurement Protocol to send data from any device that can make HTTP requests. Just use your API key and send the mention details via a POST request.
The error likely means your API key is invalid or the payload format is incorrect. Check your API key in account settings and verify the request body matches the documentation.
It's possible you're sending to the wrong endpoint or using an expired API key. Verify the endpoint URL and regenerate your API key if needed. Data should appear within minutes.