You attach a PDF, wait for the upload, then ask a question. The assistant says it cannot read the file, answers vaguely or misses a section you can clearly see. These are different problems, even if they feel the same. Separating the upload from the reading task helps you choose a useful next step.
First check what actually failed
Did the file fail to attach, or did it attach successfully and produce a poor answer? An upload error concerns getting the file into the tool. A reading problem concerns what the tool can extract or understand afterwards. A visible filename is a useful clue, but it does not prove that every page was processed correctly.
Copy the exact error wording before changing anything. “Unsupported file”, “limit reached” and “permission denied” suggest different causes. Keep the original file unchanged while you experiment with a copy.
Check the tool’s current requirements
Supported formats and limits differ between tools, accounts and features. Google’s Gemini help page distinguishes upload allowances, storage limits and files that contain too much information for good results. It also notes that organisational settings can affect access to files from Google Drive. Check the help page for your particular service rather than assuming all assistants work alike.
Source: Google’s file-upload requirements and error explanations
- Format: is this kind of document supported?
- Size: is the individual file within the current limit?
- Allowance: have you reached a temporary upload or analysis limit?
- Access: can your account open the file, and is the connection allowed?
Do not pay for an upgrade before identifying the cause. A larger allowance will not repair a corrupted document or make a restricted work file appropriate to share. If an employer manages the account, ask the administrator about access instead of making the file public.
Try one small, non-sensitive file
Create a short practice document containing three fictional facts, such as a workshop title, a date and a start time. Upload it by itself and ask the assistant to list those facts. This gives you a simple comparison: if the small file works, the original file may need closer inspection. If neither works, check the account, service status or upload feature.
One successful test does not identify the original problem with certainty. It simply narrows the next question. Keep a note of what worked so you do not repeat the same unsuccessful steps.
Inspect the original before asking again
Open the document in its usual app. Are all pages present? Is the text sharp enough to read? Is it password-protected? Try selecting a sentence: if you can only select the whole page as an image, the document may be a scan rather than ordinary text. Some AI tools can interpret scans, but results depend on their capabilities and image quality.
For a practice scan, use a clear, straight image with the page edges visible. If you use text recognition to create a readable copy, check the converted words and numbers. An incorrectly recognised date can travel into every later answer.
Reduce the task, not just the file size
A smaller file in megabytes can still contain a great deal of text. Google warns that large uploads can lead to missed details or connections. Try the relevant section with its heading, page number and any nearby notes needed to understand it. Keep the original available for checking.
For example, if you need the setup instructions for a fictional community event, begin with that section instead of uploading a folder of unrelated material. Do not remove conditions on the next page just to make the excerpt shorter.
Check the result and know when to stop
Compare the answer with the original, especially names, amounts, dates and tables. If important material remains unreadable, use another supported method or read it directly. Repeated confident guesses are not progress.
For support, provide the error message, file type, approximate size and the result of your small-file test. Avoid posting the document publicly to demonstrate the issue. Your goal is a reliable way to work with the information, not simply getting an attachment icon to appear.

