logo
search
Others

Fix Azure Content Safety Groundedness Detection Always Returning True

Maira MehtabMaira Mehtab Sep 22, 2026 869 views

Question details

The user needs a resolution for an issue where the Azure Content Safety API groundedness detection evaluates every answer as correct, consistently returning true.

Product
Azure Content Safety API
Device & OS
not provided
Scenario
Evaluating whether an AI-generated answer is accurately grounded in source documents using the Azure Content Safety API.
Observed behavior
The groundedness detection API fails to identify ungrounded or hallucinated content, consistently returning a 'true' result for all evaluations.
Before you start

Verify your API version and check your request payload to ensure you are not passing default or malformed values that might inadvertently bypass the evaluation logic.

Solution 1Recommended

Submit the Technical Issue to Microsoft Q&A

Since this is a backend API behavior, it requires deeper technical investigation by Azure engineers. Posting on the official forums is the recommended path to resolution.

When an Azure API behaves unexpectedly and consistently returns identical responses regardless of input, it may indicate a service-side bug or a specific model configuration issue. Microsoft Q&A is the official channel where Azure support engineers monitor and investigate these anomalies.

1
Navigate to the Microsoft Q&A Forum

Open your web browser and go to the official Microsoft Q&A portal at learn.microsoft.com/en-us/answers/ or specifically the Azure tag page at learn.microsoft.com/en-us/answers/tags/133/azure.

2
Prepare Your API Details

Gather your exact request format, the API version you are calling, and your specific model configuration.

3
Collect Sample Payloads

Copy the exact response payload along with the sample source content and the ungrounded answer that triggered the false positive.

4
Post Your Question

Create a new thread detailing the issue. Include all gathered logs, sample data, and API parameters so the Microsoft support team can reproduce and investigate the behavior.

Redact Sensitive Data: Ensure you remove any authentication keys, subscription IDs, or sensitive proprietary content from your sample payloads before posting on the public forum.
Free Microsoft Office alternative

Draft and Manage Your API Documentation with WPS Office

While you wait for Azure support to resolve your API issues, you can easily track your testing logs, draft technical queries, and manage developer documentation using WPS Office. It is a robust, lightweight, and free alternative to Microsoft Office.

  1. 1. Download WPS Office: Visit the official WPS Office website and download the installer for your operating system.
  2. 2. Install the Software: Run the setup file and follow the quick on-screen instructions to install the suite.
  3. 3. Start Documenting: Open WPS Writer or Spreadsheets to begin tracking your Azure API testing logs and configurations.
Fully compatible with Microsoft Word, Excel, and PowerPoint formatsFree and lightweight, ensuring fast performance on any devicePerfect for organizing API testing logs and technical documentationBuilt-in PDF tools for easily sharing your technical reports
microsoft office alternative - wps office

Frequently Asked Questions

What is groundedness detection in Azure Content Safety?

Groundedness detection is an AI feature designed to verify if a generated response is fully supported by the provided source documents. It helps identify and mitigate AI hallucinations in applications.

Why would the groundedness API constantly return a true value?

This can happen due to a service-side bug, using an outdated API version, or formatting errors in the request payload where the source text and evaluation text are not parsed correctly by the model.

What information should I include when reporting an Azure API bug?

To get the fastest resolution from Microsoft engineers, include your API version, model configuration, complete request headers/payload (excluding secrets), the response payload, and the specific text sample used.