Imagine an AI assistant that could analyze the subtle nuances of a child's play, unlocking hidden insights into their development and communication. Researchers are exploring how large language models (LLMs), the technology behind AI chatbots, could analyze conversations between children with autism and adults during clinical observations. By examining transcripts of these interactions, LLMs can identify who is speaking, what activities are happening, and even assess the child's language skills. The research tested several leading LLMs, finding they could often outperform non-experts in analyzing these complex interactions. For example, they proved adept at identifying activities like figure play or coloring from snippets of conversation. They could even assess a child's language development stage, from pre-verbal to word combinations. However, there are challenges. LLMs sometimes make errors or “hallucinate” information, highlighting the need for further refinement. This technology holds the promise of assisting clinicians in diagnosing and supporting children with autism, offering a new lens into the complexities of neurodevelopmental differences. Future research will explore incorporating audio and video data to create a more complete picture and address potential biases in AI analysis. This is a significant step toward harnessing the power of AI to understand and support neurodiversity.