Did ChatGPT read the diagram in your Word document?

A summary of a Word handover guide could include every paragraph and still miss the pictured decision tree. If that picture explains when to call a supervisor, the missing information changes what someone does next. OpenAI's existing ChatGPT Enterprise file guide says images embedded in files other than PDFs aren't processed. It recommends converting the file to PDF before uploading. The guide also distinguishes a PDF attached in a prompt from one stored as GPT Knowledge or a Project File; those stored files use text-based retrieval. Where you put the file matters as well as its extension. Before using a summary to brief a colleague, I'd choose one decision that appears only in the diagram and compare the answer with the original. For a visual question, attach the relevant image separately, or use the documented direct PDF route in Enterprise. Then check the answer again. Converting a file gives the visual content a supported route; it doesn't prove the interpretation is right. This is a reading of the Enterprise documentation, not a file test or a claim about every ChatGPT plan. The handover guide is an example. I'd rather catch one missing instruction before sharing the brief than ask the next shift to discover it during the job.