Which tool to choose
All assistants have a cap on file size, different from one to another, but the strategy to get around it is almost always the same.
- If you need all the content of a huge file, the way is to split it into parts and upload them in sequence, working one piece at a time.
- If you need only a part (a chapter, a section, a few pages), don't upload the whole file: extract or copy only that part. It's faster and more precise, because the AI focuses on what matters.
For someone who isn't experienced the second route is almost always the best: you rarely truly need a whole document of hundreds of pages at once. Identify what you need and give the AI only that.
How to do it
From a computer or a phone the principle is identical: reduce what you give the AI to a digestible size.
- Decide whether you need the whole file or just a part. It's the choice that changes everything else.
- If you need only a part, open it and copy it. For a PDF: open the document, go to the pages you're interested in, select the text and copy it. Then paste it into the chat. The operating syntax:
The complete file is too big to upload. I'm pasting only
the part I need. Work on this as if it were the document:
[paste the section you're interested in here]
- If you need everything, split the file. Search for "split PDF online" (a free service that breaks a PDF into several smaller files), split it into parts, and upload them one at a time telling the AI they're pieces of the same document.
- If it's an image that's too heavy, compress it or resize it with a free service (search for "compress image"), keeping it sharp enough if it contains text to read.
- When you upload in pieces, keep the thread. At the end of each part write: "this was part 1 of 3, wait for the others before concluding". That way the AI doesn't close the reasoning on the first piece.
- Check: after the last piece, ask for an overall check: "now you have all three parts, give me a summary that holds them together". If it cites elements from all the parts, it has connected the whole.
A concrete example
Chiara has to get the AI to analyze a three-hundred-page technical manual, but the file is huge and the upload is rejected. She realizes she only needs the chapter on maintenance, about twenty pages.
She opens the PDF, goes to that chapter, selects and copies the text, pastes it into the chat. The AI replies right away and well, because it works on twenty targeted pages instead of three hundred. The big file wasn't an obstacle to force: it was a signal that she was giving the AI much more than she needed.
When it does NOT work (and how to fix it)
If even the split parts stay too big
Split smaller. Instead of three parts of a hundred pages, make them twenty or thirty. If the weight stays high because the PDF is full of images, try to extract just the text from it (some online services do "PDF to text" or "PDF to Word") and upload that, much lighter.
If you really need everything and splitting it is too cumbersome
When the whole document matters and breaking it up by hand is impractical, consider a tool designed for long documents: some services were born to analyze entire books or archives and handle sizes that the normal chat rejects. Search for "AI to analyze long documents" and choose the one suited to your case, bearing in mind that you often pay.
If you paste the text but it's too long even that way
The chat too has a limit on how much text it handles at once. Break the pasted text into several messages, numbering them ("part 1 of 4..."), and ask the AI to wait for the last part before working. It's the same principle as splitting the file, applied to the pasted text.
If the compressed image becomes unreadable
You compressed too much. If the image contains text the AI has to read, heavy compression makes it grainy and unusable. Redo it reducing less, or crop just the portion with the text you need instead of shrinking the whole image: a sharp crop beats a whole blurry photo.
A tip from someone who actually uses it
Before fighting with a giant file, ask yourself what you really need from that file. Nine times out of ten the answer is "a part". Uploading the whole document when you're interested in a chapter isn't just harder: it also worsens the answer, because the AI gets lost among pages that aren't relevant. Giving less, but targeted, is easier to upload and makes it answer better.
Frequently asked questions
How do I split a PDF if I don't have programs?
With a free online service: search for "split PDF", upload the file, choose where to cut it or every how many pages, and download the pieces. For confidential documents, choose a service that states it deletes the files afterward, or use the splitting feature present in some PDF readers already installed on your computer.
Does uploading in pieces worsen the quality of the answer?
If you manage the thread well, no. The risk is that the AI answers on the first piece forgetting the following ones: to avoid it, say at each upload that it's a part of a whole and ask for a single summary only at the end. Done this way, working in pieces gives results comparable to the whole file.
Why is there a size limit, if the AI is so powerful?
Because reading and holding a huge document in memory takes resources, and every service sets a cap to stay fast and sustainable. The limit doesn't measure the AI's intelligence: it measures how much material it can handle at once. Getting around it by giving it smaller pieces is normal, not a trick.
Am I missing something by uploading only a part of the file?
Only if you choose the part badly, and this is the point not to underestimate. The risk isn't giving the AI little text: it's giving it the wrong piece and then trusting an answer built on incomplete information. Before extracting a section, make sure it really contains what's needed to answer your question. The right part is worth more than the whole document; the wrong part is worth less than nothing, because it gives you a self-assured answer founded on emptiness.