Which tool to choose
If your assistant reads files, upload several documents into the same conversation. If it works only on text, paste in the contents one after another, labeling each one ("Document 1: ...", "Document 2: ..."). For repeated use on the same materials, check whether there's a project space where you can keep them all available. The tool matters less than the clarity with which you distinguish the sources: the AI works well on several documents only if it understands which they are and what they should become together.
How to do it
- Bring the documents into the chat: upload or paste them, giving each a clear label.
- Say what to do with them together: "compare these two quotes," "summarize the common points of the three articles," "find the contradictions among these notes."
- Ask, where needed, that it indicate which document each piece of information comes from, so you can verify.
- Dig deeper: once you have the synthesis, ask targeted questions about a single document or a specific comparison.
A concrete example
Marta had to choose among three insurance offers, each written differently and hard to compare. She uploaded the three documents, labeled them Offer A, B and C, and asked: compare them on price, coverage and deductibles, and tell me the differences and weak points of each. The AI placed the three offers side by side on the criteria that mattered, highlighting where one was more expensive but more complete and where another hid a high deductible. In ten minutes she understood a choice she would have put off for days on her own.
When it does NOT work (and how to fix it)
If the AI confuses the documents with each other
It happens when they aren't well distinguished. Label them clearly and ask it to always cite the source of every statement ("according to Document 2..."). That way you follow where each thing comes from and notice if it's mixing them.
If it favors one document and ignores the others
It can happen that it focuses on the first or the longest one. Call it back: "consider all three equally" and, if needed, ask questions separately document by document before asking for the overall synthesis.
If the documents are too many or too long
Beyond a certain amount the AI struggles to hold them all together. Work in smaller groups, or first make a summary of each and then compare the summaries. Splitting the work prevents it losing pieces drowned in the mass.
A tip from someone who really uses it
Be explicit about the verb: compare, merge, find the differences, synthesize are not the same thing, and the AI does exactly what you ask. "Look at these documents" is too vague and you get a generic summary; "compare A and B on price and list where they differ" gets you exactly what you need. When the information is scattered across several sources, the value of the AI is cross-referencing them for you — but only if you tell it how to cross-reference them. Before uploading the files, ask yourself what you really want to come out of it, and say it with the right verb: it's the difference between a useful answer and a summary you could have made yourself.
Frequently asked questions
How many documents can I give at once?
It depends on the assistant and the length of the files. A few well-distinguished documents are handled better than many huge files: if there are too many, it's best to work in groups or summarize each one first.
How do I keep it from mixing the sources?
By labeling them clearly and asking it to cite which document it takes each piece of information from. The labels and the citations let you follow the thread and unmask any confusion.
Can I have it compare a document of mine with information from the web?
If your assistant searches online, yes: you give it your document and ask it to compare it with what it finds on the net. Remember to have it cite the web sources and verify them, as always for current matters.
If I upload several files, does the AI definitely consider them all?
No, and it's good to know so you don't trust the first synthesis. With several documents the AI may focus on one, give less weight to another or confuse them, especially if they're long or poorly distinguished. The answer looks complete but may have overlooked a source. That's why it's best to label the documents well, ask it to cite where each piece of information comes from and, on the important points, verify that every file was actually considered. Taking it for granted that the AI weighs all the sources equally is the most common way to base a decision on a synthesis that was missing one.