External API result JSON is being interpreted

I have a Custom GPT that talks to an external API to fetch my data and then I ask it to analyze and provide recommendations. The issue I am having it that each message/turn is only able to make only one API call and I need to make a few to find the data. But, then by the time we get to message 2 or 3, GPT starts summarizing or even inventing/guessing the data. It claims that it’s a limitation imposed by OpenAI to protect my data. Is there any way for the conversation to retain and use the API results verbatim?