Last reviewed
Correct answer: B. Outside Enterprise, document retrieval is text-based, so embedded images are discarded
Explanation
The principle — how much of a document reaches the assistant depends on the retrieval method, and that varies by plan and file type.
Why the key is correct — OpenAI states that ChatGPT Enterprise supports Visual Retrieval for PDF files, and that all other plans and document files only support text-based retrieval, which means ChatGPT will extract digital text from the file and discard any images. A chart that exists only as a picture is therefore not there at all.
Why the others are wrong — a file over the limit is refused rather than partly read, and the text was read fine here. Nothing is skipped for looking decorative. And images inside documents are not universally unreadable; the constraint is the plan's retrieval method.
Remember this — the fix is to attach the chart as its own image, which is read directly. Worth checking whenever an answer quietly omits something you know is in the file.
Sources — OpenAI's File Uploads FAQ.
Sources
“ChatGPT Enterprise supports Visual Retrieval for PDF files. All other plans and document files only support text-based retrieval.”
“This means that ChatGPT will extract digital text from the file and discard any images.”
Practise 10 questions on this topic
Take ChatGPT Basics — Timed Test 1 (10 questions) — scored instantly, explanation for every question, no login.