How to solve wrong citations in azure ai agents with bing grounding

Question

Description:

When asking question which require multiple web searches the citations in the answer only fit one of the web searches. Is there any fix to this issue?

Example:

Question:

'attractions in Berlin and Frankfurt'

Answer:

'In Berlin, some of the top attractions include the Brandenburg Gate, a neoclassical monument symbolizing peace and unity; the Reichstag Building, which houses the German parliament with a spectacular glass dome; and the Berlin

Wall Memorial, which offers insights into the historical divide of the city. Additionally, Museum Island is a UNESCO World Heritage site offering collections of art and history【5†source】.

In Frankfurt, the Römer is a notable historic city hall located in a scenic old town area. The Städel Museum is renowned for its extensive art collection ranging from the Old Masters to contemporary works. Other attractions include the towering Main Tower for panoramic city views, the Palmengarten botanical gardens, and the historic Kleinmarkthalle market【6†source】.'

Citations:

• text='【6†source】'. url='https://www.thebrokebackpacker.com/best-places-to-visit-in-frankfurt/', title='10 BEST Places to Visit in Frankfurt (2025) - The Broke Backpacker'
• text='【5†source】'. url='https://www.phonesimgo.com/blog/Frankfurt-2025-Must-see-attractions-and-practical-advice', title='Frankfurt 2025: Must-see attractions and practical advice

Finding:

The source for berlins attractions does not contain any information about attractions in berlin.

All sources only contain information about Frankfurt.

Answer 1

Hi there Uhlmann, Marcel

thanks for using QandA platform

My suggestions is to , refine prompts (e.g., request separate sources for each city), adjust retrieval settings, force additional searches per entity, or implement post-processing to verify citations. If the issue try giving the feedback direclty to ai studio.

If this helps kindly accept the answer thanks much.