Why we built CaseSharp.
The long version, with citations. Five short essays on Australian legal research, the limits of general AI in this domain, and what research depth ought to cost a Sydney solicitor in 2026.
A law student in 2026 has four real options, and all four fail at different points in the week.
Before we explain what CaseSharp is, it is worth being precise about what it is competing with. The Australian student research stack has not been stable for years and the pieces a student relies on today are either institutional, friction-heavy, or architecturally shallow.
Institutional premium databases
Ends on graduation day
The university-funded subscriptions that carried you through every problem question, every moot, every case-note assignment. The day your university login stops working, that access stops too. What you graduate into is not the research tool you graduated with.
Subscription professional research tools
Around A$95 per month and up
Built for barristers and solicitors with firm budgets. Strong on citation tracing and primary material. Pricing is set above what a third-year student or a newly admitted solicitor can comfortably justify out of their own pocket. Student access programs exist on some products but typically require emailing editors and waiting.
Free public databases
Free · search language from 2005
Hundreds of databases. Every Australian law school teaches them. The public floor. Boolean search with no natural-language layer, no summaries, no citation graph visible to the reader, no treatment classification. You can find a case. Understanding how that case has been treated since is a separate project.
AI-native challengers and generic chat
Modern UX · variable grounding
Fast interfaces. Broad corpora. Architecturally many of them are chat wrappers over a case index, without explicit citation graphs, canonical authority identities, or paragraph-level treatment. Useful for a first pass. Not the shape of a serious research product.
CaseSharp does not replace any of them by itself. It sits in the middle: grounded AI explanations against verified paragraph-level source text, a real citation graph, treatment classification, AGLC4 by default, Australian residency, and pricing a third-year law student can actually pay.
Australian law is not a search problem. It is a treatment problem.
Search returns a list of cases. Treatment tells you whether each case is still good law. Most legal-research tools, including the free ones every Australian law school teaches, were built around the search problem. They are excellent at it. They tell you almost nothing about treatment, and what they do tell you is at the case level: this case has been cited 213 times. They do not tell you which paragraph was cited, by which later court, applying or distinguishing or refusing to follow.
CaseSharp was built around treatment. Every authority in CaseSharp carries paragraph-level citation tracing. Open a judgment and the right column shows every later case that cited each numbered paragraph, with the treatment classification visible at a glance. The treatment status of the authority as a whole is the headline. This is not a feature. It is the thesis of the product.
Sharp will say it does not know. It will not invent a citation.
Every Sharp answer is grounded in the verified text of judgments in the corpus. When the corpus contains a clear answer, Sharp produces it with inline paragraph citations you can click through to read. When the corpus does not contain a clear answer, Sharp says so. When a question is outside Australian legal research, Sharp declines.
There is no hallucinated authority in CaseSharp. There is no decorative citation. The bar a Sydney solicitor would set for their own research is the bar Sharp is held to. If we cannot meet it on a given question, we tell you. The architecture supporting that promise is the work, not the marketing copy.
The single largest involuntary downgrade in Australian legal practice.
Roughly 7,500 Australian law students graduate each year. The day they do, their institutional research-database access ends. What replaces it, for most of them, is materially worse than what they had. The cliff is documented, predictable, and unsolved by anyone who pays attention to whether new graduates have access to the same research depth they trained on.
CaseSharp Free is not a trial. It is a lifetime tier with three Sharp chats and one brief, on the full Reader and full corpus. A graduate who signs up the day they finish their last exam keeps an account that does not expire. When their work demands more, Pro is A$49 per month with no contract. The pipeline is the same engine into a surface designed for how new graduates actually work.
Four things changed in the last eighteen months that made this product buildable.
Editorial labour cost has collapsed
The human editorial teams that priced premium research at institutional rates are no longer the only way to produce headnotes, treatment labels, and practice points. A corpus the size of the major Australian case-law collection is now editable at under A$50,000 per year of compute rather than A$4M per year of payroll.
The market is consolidating around bundled AI
The enterprise and PMS-bundled answers are locking in over the next eighteen months. An independent student-first research tool that stays with graduates into practice has a narrow window to establish habits first.
Australian regulators are naming the pattern
In 2025 the Victorian Legal Services Board and Commissioner sanctioned a solicitor for submitting AI-hallucinated case citations in the Federal Circuit and Family Court of Australia. Student habits formed now on grounded, verification-first tools are the direct answer to the risk the profession is starting to see.
Capable reasoning models are production-ready
Thinking-token reasoning, structured JSON output, retrieval-before-generation, and paragraph-level pinpointing are now table stakes. A student research tool that treats these as defaults rather than experiments is able to ship safely today. Building without them in 2026 is negligent.
The four pieces that shaped how CaseSharp was built.
The Australian legal-AI market, the economics of editorial labour, the graduation cliff, and the safety architecture for citation-grounded AI work. If you are evaluating CaseSharp seriously, these are the pieces that frame the product.
Strategy
The graduation cliff
The involuntary downgrade every Australian law graduate faces on the day their university research-database access ends, and the student-to-firm pipeline CaseSharp was designed around.
Read →
Methodology
Paragraph-level treatment, explained
Why treating citations at the paragraph level changes which case answers your problem, and how the difference plays out in litigation, academia, and any setting where the question is what was followed, not just what was cited.
Read →
Practical
AGLC4 mistakes that cost marks
A field guide to the small punctuation, italics, and pinpoint conventions that actually matter, written by people who have read every section of the rulebook so that you do not have to.
Read →
Reference
Understanding case treatment
Followed, applied, distinguished, overruled. What each treatment classification means in Australian appellate practice, and how CaseSharp surfaces the difference.
Read →