Westlaw's Real Moat Isn't the Case Law. It's the Way the Law Is Filed.
Westlaw's Legal Professionals segment runs at about 97% recurring revenue and grows 8% organically — not because its search is unbeatable, but because a century-old classification system and a 2025 copyright ruling make leaving prohibitively expensive.
Comes with a free Switching-Cost Ledger template.
A judge's opinion is, legally, free. It is government text, written by a public official, owned by no one — you can photocopy it, post it, sell it, quote it forever. So you would think the company sitting on a near-permanent grip over American legal research would be sitting on something scarce. It isn't. Westlaw's most valuable asset was never the law. It was the index.
The official story is that Thomson Reuters owns the cases. It sells access to court opinions. What it actually sells is the editorial layer wrapped around those opinions — the short summaries of each legal point, and the century-old system that decides which drawer every point gets filed in. The opinions are free. The filing cabinet is not, and everyone in the profession has memorized where the drawers are.
That is the thesis, and it is sharper than the usual 'big incumbent has scale' line: Westlaw's lock-in is structural, not commercial. Thomson Reuters does not need to win on product merit. It needs only for the way lawyers were trained to find the law to keep belonging to it — and that belonging is now backed by copyright law itself.
What Westlaw actually sold in 1975 — and it wasn't the cases
The origin myth is that Westlaw launched in 1975 as a full-text database of American case law. It didn't. When West Publishing put Westlaw live at Case Western Reserve Law School that year, what it offered was access to its headnotes — West's own editorial summaries of the legal points in each case — and nothing else.910 You could not read the opinion itself. Full-text retrieval wasn't completed until 1978.11 For its first years, Westlaw was the editorial layer, with no underlying text at all.
That sequence is the whole tell. West did not get rich by being first to digitize public documents. It got rich by being the place where a generation of lawyers learned to navigate the law — through headnotes filed into West's proprietary classification scheme. Teach an entire profession to think in your taxonomy, and you have built something no rival can simply copy by scanning the same court records. The cases are commodity. The map of the cases is not.
“Westlaw's 1975 debut at Case Western Reserve provided access only to West's editorial headnotes — not the full text of cases. Full-text retrieval was not completed until 1978.”5
How a $3.43 billion bet became a 97%-retention machine
Thomson Corporation saw what the editorial layer was worth. In 1996 it bought West Publishing — Westlaw's creator — for $3.43 billion in cash, a deal that closed in 1997 after a Department of Justice antitrust review.4 Note who did the buying: not 'Thomson Reuters,' which didn't exist yet. Thomson would not merge with Reuters until 2008, twelve years later.4 The popular shorthand collapses that history, but the timing matters, because the asset Thomson paid for was already mature: a self-reinforcing position built over decades, not a tech bet on the future.
The payoff shows up today as numbers that read like a utility, not a publisher. In full-year 2023, the Legal Professionals segment generated $2.807 billion in revenue — roughly 41% of the entire company — of which 95% was recurring.1 By the first quarter of 2024, recurring revenue had climbed to 97% of the segment, growing 8% organically at an adjusted EBITDA margin of 47.4%.2 Read that pair again: nearly all revenue is recurring — locked in before the year begins — and the segment still grows. That is not what a competitive market looks like. That is what a switching cost looks like.
| The court opinion | West's editorial layer | |
|---|---|---|
| Who created it | A public official | West's editors |
| Copyright status | Public domain — free to anyone | Proprietary, copyright-enforced |
| What lawyers were trained on | The raw text | The headnotes and the classification system |
| Where the recurring revenue sits | Almost none | Nearly all of it |
The 2025 ruling that turned a habit into a wall
For years the lock-in rested on habit and contracts. Then a startup tried to route around it, and the courts turned the habit into something harder. ROSS Intelligence set out to build an AI legal-research tool — and, the court found, trained it on Westlaw's headnotes to do so. On February 11, 2025, a federal court in Delaware ruled on summary judgment that ROSS's use of those headnotes infringed Thomson Reuters' copyright and was not protected by fair use, with ROSS found to have directly infringed 2,243 of the 2,830 headnotes at issue.7 It was the first U.S. ruling to reject a fair-use defense for AI training data.12
Separately, the same litigation gutted the obvious counterattack. ROSS had argued that Thomson Reuters was abusing a dominant position — refusing to license its content, tying its products together. The court dismissed those antitrust counterclaims, finding that Thomson Reuters' conduct amounted to legitimate business decisions, not unlawful behavior.8 So the moat was upheld on two fronts at once: you may not copy the editorial layer to compete, and refusing to sell it to you isn't a crime. The map is private property, and the cartographer has no duty to share it.
Isn't the AI era about to blow the doors off?
The honest objection is that all of this describes a fortress built for the last war. Generative AI can read raw court opinions — the free part — and write its own summaries, its own classifications, on the fly. If a model can do for nothing what West's editors did over a century, the proprietary layer stops being scarce, and the lock-in melts. That is the real threat, and it is not theoretical: ROSS existed precisely because someone believed it.
But watch what Thomson Reuters did with the threat. Weeks before the ROSS ruling momentum built, it spent $650 million in cash to acquire Casetext, maker of the GPT-4-powered CoCounsel legal assistant, signing in June 2023.6 The incumbent didn't fight AI — it bought the best independent AI tool and pointed it at its own proprietary content. So the AI that threatens to commoditize legal research now runs, in its strongest commercial form, on top of the very database that owns the editorial layer. And the 2025 ruling means rivals can't legally train on that layer to catch up. The fair counter is that the copyright win is on interlocutory appeal to the Third Circuit, accepted in mid-2025, and is not yet final7 — a reversal would reopen the door. But even a loss in court wouldn't undo the deeper fact: AI still needs structured, trustworthy, citable legal knowledge, and Thomson Reuters spent a century building the only library of it lawyers were trained to trust.
When the underlying content is free or commodity — public records, court opinions, raw data — the durable position is rarely in owning more of it. It's in owning the layer that makes it usable: the taxonomy, the summaries, the way an entire profession was trained to navigate it. That layer compounds in two ways the raw content can't. It's copyrightable (your editorial work, unlike the public document, is yours to defend in court), and it's sticky (people who learned to think in your categories don't switch easily). Two cautions. First, the layer only stays valuable while it remains the trusted standard — buy or build the next interface, as Thomson Reuters did with Casetext, before a challenger makes one people prefer. Second, a lock this complete invites antitrust scrutiny; defend it with genuine editorial value, not just by refusing to sell.
Westlaw's genius was never the cases. Anyone can have the cases — they cost nothing, by law. The genius was deciding, in 1975, to sell the summaries first and the text later, and to make the summaries so good that a profession built its working memory around them. A century on, that bet renews itself at 97% a year, grows while it renews, and now carries a federal court's signature on the deed. The law is public. The way the law is filed is private — and that one distinction sits at the center of a $2.8 billion-a-year segment.1
Other businesses that own the index, not the inventory
Switching-Cost Ledger
A worksheet that prices the exit. It itemizes every cost a customer eats to switch away — the contract penalties, the re-training, the data migration, the muscle memory — so you can see whether lock-in is real or just inertia waiting to break. Blank to audit your own stickiness; filled as the worked example tallying the switching costs the story's customers face.
The worked example unlocks with a subscription. See plans →
Sources
Where this comes from — the filings, records, and reporting behind it.
- 1Thomson Reuters Legal Professionals segment generated full-year 2023 revenue of $2.807 billion (recurring revenues $2.674B, transactions $133M), representing approximately 41% of total company revenue; recurring revenues were 95% of segment total.
- 2Thomson Reuters Legal Professionals segment Q1 2024 recurring revenues were 97% of total segment revenue, with 8% organic growth; adjusted EBITDA margin was 47.4%, up from 44.6% year-over-year — confirming the segment's high-recurring, high-margin profile.
- 3Full-year 2024: Thomson Reuters total company organic revenue grew 7%; 'Big 3' segments (Legal Professionals, Corporates, Tax & Accounting Professionals) grew 8% organically; 2025 outlook is 7.0–7.5% organic growth with ~39% adjusted EBITDA margin.
- 4Westlaw was created in 1975 by West Publishing; West was acquired by Thomson Corporation for $3.43 billion in February 1996 (closed June 1997 after DOJ antitrust review); Thomson did not merge with Reuters to form Thomson Reuters until 2008.
- 5Westlaw's 1975 commercial debut at Case Western Reserve Law School provided access only to headnotes from West's National Reporter System, NOT full-text cases; full-text retrieval was not completed until January 1978.
- 6Thomson Reuters signed a definitive agreement to acquire Casetext (maker of CoCounsel AI legal assistant, powered by GPT-4) for $650 million cash on June 26, 2023; the acquisition closed in August 2023.
- 7On February 11, 2025, U.S. District Court for Delaware (Judge Bibas) ruled on summary judgment that ROSS Intelligence's use of Westlaw headnotes to train its AI infringed Thomson Reuters' copyright and did not qualify as fair use — the first U.S. court ruling rejecting a fair-use defense for AI training data. ROSS directly infringed 2,243 of 2,830 headnotes at issue. The ruling is on interlocutory appeal to the Third Circuit (No. 25-8018, accepted June 17, 2025).
- 8A federal court separately dismissed ROSS Intelligence's antitrust counterclaim against Thomson Reuters, finding TR's refusal to license Westlaw content and its tying conduct represented legitimate business decisions, not unlawful antitrust violations.
- 9Westlaw's original 1975 system ran on dedicated terminals that provided access to headnotes but not full case texts; full-text case searching arrived in 1978.
- 10In 1975, West introduced Westlaw in response to Lexis; unlike Lexis, which included the full text of all published cases, Westlaw provided only its case headnotes.
- 11In 1978, locked in competition with Mead Data Corp., Westlaw began to provide full-text search for cases.
- 12The February 11, 2025 Thomson Reuters v. ROSS Intelligence ruling is the first time a U.S. court reached a conclusion concerning the application of copyright's fair use doctrine to the use of works in training data for artificial intelligence.