L

Learn AI

Track progress · learn offline

Open

Lesson 4 of 9

Giving it your sources

7 min read

The AI has "read" more than you ever will — so why, when you ask about a specific document, is it so often vague or subtly wrong? And why does pasting in the very page fix it instantly?

Fuzzy memory vs the page in hand

An AI's knowledge is baked in during training — squeezed down from a huge pile of text into a kind of fuzzy, general memory. It's great for the gist, but it blurs the specifics: exact numbers, the clause in your contract, what this article actually said. Ask from that memory and you get a plausible-sounding guess. Paste the actual page into the chat, though, and the exact words are right there on the shared page — so it answers from them instead of guessing.

Answering from memory gives you a blurry guess. Putting the real source in the context window lets the AI answer from the exact words in front of it — this is called grounding, and it's far more accurate.

Paste, upload, or point

You don't have to retype anything. Every assistant lets you hand it a source directly — paste the text, attach a file or PDF, sometimes drop in a link. However it arrives, the effect is the same: the real content lands on the shared page, and the AI works from that rather than its hazy recollection. When accuracy matters, don't ask it to remember — give it the source.

You can paste text, upload a file, or share a link. All of them do one thing: put the true source in the window so the answer is grounded in it, not guessed.

Getting a vague or hedged answer about something specific? That's the tell it's working from memory. Hand it the actual document and ask again — the difference is night and day.

The gist

You want an AI to summarise the key dates in a contract you were sent. What gets you an accurate answer?

Continue in the app

Take the whole Working with AI course — tracked

Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.