Lucio Labs answers the big question, 'How comprehensive can Legal Research get?'
Darsan Guruvayurappan
Most legal research time is not spent thinking, it is spent looking.
A partner asks a question at 4pm and it sounds simple but by 6pm the associate had run a dozen searches, opened forty judgments, skimmed most of them and set aside thirty. They are still not sure the list is complete. The analysis, the part the client is paying for, has not started.
Finding a result was never the hard part with any search box returns results. The hard part is getting to the authority that answers the question, reading enough of it to understand what was decided and being sure enough to put it in front of a partner, a client or a court.
Research has also always been rationed. You stop when the time runs out, not when you have read everything that matters. At Lucio Labs we wanted to test what happens when that limit moves. How much can a research system search, screen and read on a single question and still come back while the question is fresh?
Lucio Research is our answer so far.
Built to read, not just to answer
You ask a legal question the way you would ask a colleague and choose the jurisdiction. Lucio Research then works through the steps a careful researcher would take:
Search: It looks across hundreds of legal sources for anything that could bear on the question.
Screen: It sets aside material that is only superficially relevant, such as the case that uses the right words but decides a different point.
Read: It reads the authorities that survive screening in full, not just their headnotes.
Cite: It answers with an inline citation for every proposition, so you can open the authority and check it.
Alongside the answer, the “All sources” view shows every source it considered, sorted into highly relevant, relevant, flagged and discarded. You can see what was set aside, not only what made it into the answer.
An In-depth answer to a US question. The panel on the right shows the research behind it: what was fetched, screened, read in full and cited.
The All sources view. Every source is labelled as highly relevant, relevant, flagged or discarded, so you can see what was set aside as well as what was used.
Two modes
Standard is a focused search of the core authorities and the fastest answer. It reviews around 100 sources.
In-depth explores three times as many sources and checks for gaps before it answers. It often reads across 300 to 400 sources, sometimes more than 500 and usually returns in under two minutes.
Where it earns its place
Lucio Research is optimised for case law in the US, the UK and India. These are the research tasks where it saves lawyers the most time:
The first pass on an unfamiliar question. A question lands in an area you have not worked in for years. Standard mode gives you the core authorities quickly, so you know the shape of the law before you decide how far to go.
Questions with no single leading case. How have the English courts drawn the line between "reasonable endeavours", "all reasonable endeavours" and "best endeavours"? The answer sits across a long run of first instance and appellate decisions. In-depth mode is built for this kind of breadth.
Questions where the courts disagree. Does Title III of the Americans with Disabilities Act reach websites with no physical location? The US federal circuits have not given one answer. You need each circuit's position, not just the one that ranks first in a search and In-depth mode reads across all of them.
A position that has moved. When the Supreme Court of India has revisited a question more than once, as it has on arbitration clauses in unstamped agreements, a keyword search mixes earlier and later decisions together. Reading across the whole line lets you see how the position developed and check which decisions are current before you rely on one.
Preparing for the other side. Before a hearing or a negotiation, ask for the authorities that cut against your client's position. It is easier to deal with a case you have already read than one you first meet in the other side's submissions.
For the technically curious
A common design for AI research is a pure agent loop. The model decides on a step, takes it, reads the result and decides again. That is flexible, but it is sequential, so every extra source adds time.
Lucio Research works differently. It uses fast decision models in place of a pure agent loop, on a new architecture built to run as much of the work in parallel as possible. That is how it can read more without making you wait longer.
Why speed matters and why it is not the point
For context: in our testing, a deep research query in ChatGPT takes around 10 minutes and public models usually go through no more than 25 sources.
But speed is the easiest thing to measure, so it gets talked about first. A fast answer built on eight sources is not research. It is a guess with footnotes.
What matters is reaching the right authority quickly and being able to stand behind it. That is why Lucio Research is built to go wide before it goes fast. The time it gives back is time spent searching. The thinking stays with you: which authority is closest on the facts, whether the reasoning holds, what it means for this client.
Where it is strongest today
Lucio Research is optimised for case law research across the US, the UK and India. We are still improving its analysis of regulation and legislation, so treat questions that turn mainly on regulatory material as early work.
Lucio Research sits behind the same SOC 2 Type II and ISO 27001:2022 certification as the rest of Lucio. No customer data is used to train models.
An experiment, in the open
Lucio Labs is where we test ideas before they reach the main product. The question behind Lucio Research, how comprehensive research can be, can only be answered on real questions, by the lawyers who research them.
So for the next month, we are opening it up as a free beta. Anyone can use it, with up to 25 searches each. That is enough to put it through a handful of questions you know well.
We want to know where it falls short: the authority it missed, the case it should have set aside, the question it misread. What we learn will shape how research works inside Lucio.
The fairest test is a question you researched recently, where you already know the answer. Run it through Lucio Research. See what it finds, what it cites, what it discards and how long it takes.
Try Lucio Research: https://research.lucioai.com/
