Capabilities & use
How do you write effective prompts for legal AI?
Be specific about the task, the jurisdiction and the output you want, give the AI the source material to work from, and ask it to cite its sources. Good legal prompting looks less like a search query and more like instructing a capable junior: state the matter type and jurisdiction, attach the relevant documents, define the deliverable, and ask for authority on every point. Then verify what comes back.
Give context, not keywords
State the jurisdiction, the matter type and your position. AI works far better with context than with a bare keyword search — tell it what you are doing and for whom, not just the topic.
Ground it in your documents
Attach the contract, the data room or the precedents. AI grounded in your own source material is more accurate and more useful than AI answering from general memory, and it keeps the work anchored to your matter.
Ask for sources, then verify
Instruct it to cite authority for every claim, and then check those citations. The prompt sets the expectation; your verification confirms it. With purpose-built legal AI, a citation on every answer should be the default rather than something you have to request.
Iterate
Refine with follow-ups — narrow the jurisdiction, change the output format, ask it to focus on one clause — rather than expecting a single perfect prompt to do everything.
Frequently asked questions
Do lawyers need to learn prompt engineering?
Not formally. Clear instructions — task, jurisdiction, documents, desired output, and a request for sources — capture most of the value, and purpose-built legal AI needs less prompting than a general chatbot.
Why does legal AI give me vague answers?
Usually too little context. Add the jurisdiction, the relevant documents and the specific deliverable you want, and ask for citations, and the answers sharpen.
Should I always ask the AI to cite its sources?
Yes — ask for authority on every point and then verify it. Good legal AI provides a source by default, but asking makes the expectation explicit.
See how Quillio handles this in practice
AI built for Australian and New Zealand law — a citation on every answer, client content stored in Australia, and a free trial so you can test it on your own files.