Why Caselaw Citations Still Depend on Dead Print Reporters

· LawDiver

American lawyers still cite volume and page of commercial print reporters even as those books die. Copyright fights, the Thomson/West antitrust decree, the end of the Federal Appendix, Florida's So. 3d rules, and what free caselaw data still gets wrong about citations.

American lawyers identify cases the way printers once stocked shelves: volume, reporter abbreviation, page. That string — 531 U.S. 98, 799 F.2d 1219, 772 So. 2d 1243 — is not a neutral government ID. It is a location in a commercially arranged book, or a digital shadow of one. Courts still require those strings in briefs. Libraries cancel the bound sets. Publishers discontinue entire series. Open databases often ship the opinion text without the reporter cite a filing rule demands.

This article is a source-linked history of how that mismatch happened — copyright doctrine, an antitrust settlement, the quiet death of print, Florida's citation rules as a live stress test, and measured citation-coverage gaps — plus what it means for anyone building on free caselaw data. We build LawDiver and CiteDiver, so we are not neutral. Empirical claims that are ours rather than the public record are disclosed and reproducible at the end.


On this page

  1. The problem in one sentence
  2. Reporter timeline: dates that still run the cite
  3. Copyright: who owns the page numbers?
  4. Antitrust: Thomson buys West
  5. The fight moves to AI training
  6. Print status: when the books stop
  7. Citation formats: Florida as the stress test
  8. Florida citation coverage charts
  9. Florida So. 3d reconstruction (LawDiver pipeline)
  10. Data and market: open corpora, consolidating platforms
  11. What free data still often lacks
  12. Method and bias disclosure
  13. Frequently asked questions
  14. Sources
  15. Related reading

The problem in one sentence

Judicial opinions are public; the citation infrastructure lawyers must use was built by commercial publishers and never fully replaced when print economics collapsed.

Everything below is an unpacking of that sentence.


Reporter timeline: dates that still run the cite

Before the case law, here is the address book itself — the reporter series whose volume/page pairs still appear in briefs, even when the bound volumes have stopped.

Reporter / seriesApprox. spanRole in citation practiceStatus
U.S. (United States Reports)1790–presentOfficial SCOTUS reporter; Rule 9.800 and Bluebook prefer itStill printed (bound volumes & preliminary prints)
S. Ct. / L. Ed. 2dParallel SCOTUSCommercial parallels used with or without U.S.Ongoing commercial
F. / F.2d / F.3d / F.4th1880–present (series succession)Precedential federal appellateOngoing (F.4th current)
F. App'x (Federal Appendix)2001–2021 (861 vols)Non-precedential federal appellateDiscontinued — no F. App'x 2d
F. Supp. / F. Supp. 2d / F. Supp. 3dDistrict courtFederal trialOngoing
So. / So. 2d / So. 3dSouthern regionalFlorida (and AL, LA, MS) filing cites under Fla. R. App. P. 9.800So. 3d current; So. 2d closed for new Florida opinions
Fla. L. Weekly (FLW)Advance sheetsInterim Florida cite until Southern assignmentOngoing commercial (FLW)
WL / LEXIS proprietaryOnline-onlyFallback when no reporter/FLW cite existsOngoing databases
Neutral / public-domain citesVaries by stateVendor- and medium-neutral formsAdopted in some jurisdictions; not Florida's primary form for Southern-reported cases

The practical punchline: discontinuing a reporter does not repeal the citation rule that named it. Florida Rule 9.800 still lists Federal Appendix as a fallback for unpublished federal appellate decisions — a closed set after 2021. Florida still requires So. 3d for ordinary state appellate opinions even as libraries cancel regional reporters (UCI Law cancellations example).


For most of the twentieth century, West Publishing's National Reporter System was the de facto address book of American case law. Competitors who wanted to let users cite "to West" without owning West's books needed star pagination — inline markers showing where West's pages broke. West treated that pagination (and much of its editorial apparatus) as proprietary.

Mead: the Eighth Circuit protects pagination (1986)

In West Publishing Co. v. Mead Data Central, Inc., 799 F.2d 1219 (8th Cir. 1986), the Eighth Circuit affirmed a preliminary injunction against LEXIS star pagination. The court treated West's arrangement of cases as a copyrightable compilation and held that comprehensive use of West page numbers would appropriate that arrangement. For a decade, that decision chilled competitors who wanted West-compatible pinpoint cites without a license.

Matthew Bender: Feist arrives at the reporters (1998)

After Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340 (1991), originality — not sweat of the brow — became the constitutional floor for copyright in factual compilations. The Second Circuit applied that floor to West in twin 1998 opinions:

Those holdings did not abolish commercial reporters. They did mean that, in the Second Circuit's view, the text of opinions plus page-locator stars is not a West monopoly. The industry still settled and licensed; doctrine and dealmaking ran on parallel tracks.

Georgia v. Public.Resource.Org: government edicts (2020)

Georgia v. Public.Resource.Org, Inc., 590 U.S. 255 (2020) held that annotations in Georgia's Official Code, drafted by a state commission as part of the official code, fall under the government edicts doctrine and are not copyrightable. The opinion is about official annotated codes, not National Reporter volumes — but it reinforced a broader public-access principle: when the state speaks as lawgiver, copyright is the wrong tool to meter access. For caselaw builders, the practical takeaway sits next to Matthew Bender: the law's text is free; the commercial scaffolding around it is where the fights continue.


Antitrust: Thomson buys West

Copyright was only half the story. In 1996 the Department of Justice and several states challenged Thomson Corporation's acquisition of West Publishing. The competitive concern was straightforward: combining major legal-publishing assets concentrated control over primary law products, secondary materials, and the citation conventions the bar already used.

Primary documents still worth reading in full:

The decree required product divestitures and licensing relief — including access to star pagination — aimed at preserving rivals' ability to compete with West-compatible citations. It did not invent a government-issued neutral citation for every U.S. court. The market kept the West address book; antitrust tried to keep the address book from becoming an exclusive lock.

That bargain still shapes today's landscape: competitors can often display West-style locators, but the assignment of volume and page for new decisions remains a publishing act, not a court clerk's automatic public ID in most jurisdictions.


The fight moves to AI training: Thomson Reuters v. ROSS

Pagination fights were about locating public opinions. The next copyright frontier is editorial work product used to train competing research systems.

Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc., No. 1:20-cv-613 (D. Del.), is the leading U.S. case on whether using Westlaw headnotes to train a competing legal-search tool is fair use. Judge Stephanos Bibas (sitting by designation) has issued multiple opinions; the fair-use fight is now on interlocutory appeal in the Third Circuit (No. 25-2153, after §1292(b) petition No. 25-8018).

Procedural arc

  1. September 25, 2023 — Memorandum opinion largely denying summary judgment on copyright, fair use, and related issues because material facts remained disputed (694 F. Supp. 3d 467).
  2. October 2024 renewals — After continued discovery and a continued trial date, the parties renewed summary-judgment motions on copyright infringement and fair use (Thomson Reuters fair-use MSJ, Thomson Reuters copyright MSJ, ROSS fair-use MSJ, ROSS copyright MSJ).
  3. February 11, 2025 — Revised memorandum opinion granting Thomson Reuters partial summary judgment on direct copyright infringement (for thousands of identified headnotes) and rejecting ROSS’s fair-use defense (Feb. 11, 2025 opinion).
  4. §1292(b) certification — The district court certified controlling questions on headnote originality and fair use for interlocutory appeal (certification opinion; petition-stage amicus: Frye).
  5. Third Circuit briefing — Merits appeal No. 25-2153: ROSS as appellant, Thomson Reuters as appellee, with extensive amicus participation on both sides.

What the parties argue

Thomson Reuters / West argue that Westlaw headnotes and the Key Number System are original editorial expression; that ROSS’s bulk-memo pipeline copied protectable headnote text to build a competing research product; and that the use is not transformative fair use because it substitutes for licensed editorial content in the same market (legal research tools), even though the underlying judicial opinions are public domain.

ROSS argues that any protectable expression in headnotes is thin; that training a non-generative retrieval model on headnotes (to find cases, not to emit headnote prose) is transformative; that market harm must be tied to a cognizable copyright market rather than competition with a data product; and that treating headnotes as an exclusive training set would lock up functional legal taxonomy.

Amici supporting ROSS (libraries, EFF-led groups, copyright professors, legal-tech platforms, and others) emphasize public access, thin copyright in legal taxonomy, and innovation for non-generative AI. Amici supporting Thomson Reuters (publishers, entertainment companies, Copyright Alliance, and others) emphasize incentives for original editorial investment and warn against treating commercial databases as free training corpora. Primary materials on LawDiver:

Secondary coverage of the amicus wave: IPWatchdog (Oct. 2025). The Third Circuit has not issued a merits decision as of this writing. Whatever the appellate outcome, the district court already signals that headnotes and other publisher-authored legal analysis are not a free training set, even when the underlying opinions are public domain. That is a different axis from Matthew Bender’s star-pagination holding — and it matters for every legal-AI product that wants editorial signal without a license.

For why probabilistic models remain unsafe as sole cite checkers even when trained on clean data, see Why do large language models hallucinate?.


Citation rules assume a living reporter series. Several of those series are dead or shrinking.

Federal Appendix (2001–2021)

West launched the Federal Appendix in 2001 to print non-precedential federal appellate decisions — the oxymoronic "published unpublished" opinions. After Federal Rule of Appellate Procedure 32.1 (effective 2007) liberalized citation of unpublished dispositions, the series had a clearer use case. Thomson Reuters discontinued it in 2021 after 861 volumes. There will never be an F. App'x 2d.

MarkerDetail
First volume / citeArbelaez v. Newcomb, 1 F. App'x 1 (D.C. Cir. 2001)
Last volume861
Last-by-page cite (often cited)Vittetoe v. Blount County, 861 F. App'x 843 (6th Cir. June 17, 2021)
Series successionNone — no F. App'x 2d

Obituaries from practitioners and librarians:

The opinions still exist online. What died was a canonical volume/page address that brief writers preferred to raw WL or LEXIS strings. Florida's own appellate citation rule still lists Federal Appendix as a fallback for unpublished federal appellate decisions — a rule that now points at a closed set.

Official reporters that still print — and libraries that stop buying

The Supreme Court continues to issue United States Reports bound volumes and preliminary prints. That is the exception that proves the trend: a government-published official reporter with a stable citation form.

Elsewhere, academic law libraries document serial cancellations as budgets collide with publisher pricing — see, for example, UCI Law Library collection budget cancellations. When the bound Southern Reporter leaves the shelf, the So. 3d cite does not become optional. It becomes a cite to a book the local library may no longer hold.


Citation formats: Florida as the stress test

Florida is the clearest working example of a jurisdiction that still requires commercial regional-reporter citations while living in a mostly digital filing world.

What Rule 9.800 actually requires

Fla. R. App. P. 9.800 (Uniform Citation System) tells Florida courts and filers how to cite. The hierarchy for Florida appellate opinions is not optional fashion — it is rule text:

AuthorityPreferred cite form under Rule 9.800
Florida Supreme Court (1887–present), in SouthernFenelon v. State, 594 So. 2d 292 (Fla. 1992)
Florida DCA, in Southerne.g. Buncayo v. Dribin, 533 So. 2d 935 (Fla. 3d DCA 1988)
Not (yet) in SouthernFlorida Law Weekly (e.g. 17 Fla. L. Weekly S42)
Not in FLWSlip opinion / docket form; optional Westlaw or LEXIS parallel
Unpublished federal appellate (after F. Reporter)Federal Appendix if published there; else slip / WL / LEXIS

That hierarchy has a practical consequence: a database that has the PDF but not the So. 3d string is incomplete for Florida filing practice, even if the text is perfect.

Neutral citation was supposed to fix this

Vendor- and medium-neutral citations — year, court, sequential decision number, paragraph pinpoints — exist so courts need not wait for a commercial volume. The University of South Carolina School of Law maintains a guide to Universal Citation: adopting jurisdictions. Adoption remains patchy. Florida has not replaced So. 3d with a statewide neutral system for precedential opinions. Until it does, reconstruction and licensing fights around regional reporters stay live.

Open tooling that has to parse whatever the bar actually types includes Free Law Project's reporters-db and eyecite — the same stack much of the industry, including LawDiver, builds on.

Parallel cites are not a theoretical problem

Consider the Florida and U.S. Supreme Court decisions from the 2000 election litigation:

CaseCite stringDate filed (cluster)LawDiver
Gore v. Harris772 So. 2d 1243Dec. 8, 2000/casedisplay?opinion=109200000001666
Gore v. Harris773 So. 2d 524Dec. 22, 2000/casedisplay?opinion=109200000000879
Bush v. Gore531 U.S. 98 (and parallels)Dec. 12, 2000/casedisplay?opinion=200200000000073

A cite checker that resolves only one parallel, or that cannot map advance-sheet / FLW forms to later So. 2d/3d assignments, fails real briefs. That failure mode is infrastructure, not UX.


Florida citation coverage charts

Digitization and court-site harvest give you opinion text. Reporter assignment is a separate data problem. Florida makes the gap measurable.

CourtListener public API — Florida Supreme Court

We queried CourtListener's public v4 search API and counted the fraction of results with a non-empty citation array (method in the disclosure). Re-verification on September 11, 2026 for the Florida Supreme Court (court=fla):

YearCourtOpinions countedWith non-empty citationShare
2005Fla. (Supreme Court)327 (full year)327100%
2015Fla. (Supreme Court)222 (full year)20391.4%
2020Fla. (Supreme Court)140 (full year)00%
2024Fla. (Supreme Court)131 (full year)00%

That is the entire story in four rows. CourtListener still has the opinions. By 2020 the free graph's citation array for Florida Supreme Court opinions is empty for the whole year — while Rule 9.800 still wants a Southern Reporter string. The same pattern is worse in the District Courts of Appeal samples described in the disclosure (400-opinion draws returning zero citations). Re-run the public API recipe anytime; if Free Law Project backfills these years, we will update the table.

LawDiver corpus — Florida opinions with a Southern reporter key

Separately, here is what our own corpus shows for lead Florida opinions (Supreme Court + DCAs) that already carry a So. / So. 2d / So. 3d reporter key. This is LawDiver's live store as of September 11, 2026, including CAP-era rows and our own reconstruction work — not the CourtListener search sample above. It still shows the cliff when online publication outran reporter metadata:

YearFlorida lead opinionsWith Southern reporter keyShare
20052,2292,229100%
20102,9902,990100%
20132,1942,194100%
20143,5602,37566.7%
20154,2312,43757.6%
20189,0812,72830.0%
20202,32550721.8%
20223,09282826.8%
20244,1871,12826.9%
20253,36887526.0%

Florida Supreme Court alone stays higher longer, then slips as online-first opinions accumulate without Southern keys:

YearFla. Supreme Court leadsWith Southern keyShare
2005–2013(each year)(each year)100%
201519918291.5%
201839431780.5%
20201266450.8%
20231135246.0%
2024934346.2%
2025985859.2%

The 2014–2018 drop is the story in one chart: text coverage kept growing; Southern citation attachment did not.


Florida So. 3d reconstruction (LawDiver pipeline)

Rule 9.800 still wants So. 3d. When upstream free metadata lacks the key, LawDiver runs a deterministic reconstruction pipeline that matches Florida So. 3d assignments (volumes 276–432, filing dates roughly 2011–2026) to opinions in our corpus.

Treat the following as a pipeline claim, not a public statistic (see disclosure).

Reconstruction set coverage

MetricValue
Rows in reconstruction set16,153
Volume range276–432 So. 3d
Year span2011–2026
Fla. Supreme Court rows810
1st DCA3,203
2d DCA2,141
3d DCA4,465
4th DCA2,744
5th DCA2,369
6th DCA421

Match outcomes (name-score floor 0.85)

From the September 9, 2026 assign report against our corpus:

OutcomeCountMeaning
Already present5,121Reporter key already on the cluster
Accept (new write)458Unambiguous court + date + name match above floor
Low name score7,888Candidate(s) found but below 0.85 similarity
No candidate2,188No corpus opinion on that court/date window
Ambiguous330Multiple plausible matches
Other / short / other So. 3d168Rejected for span, conflicting key, or related reasons

A lower-floor exploration pass (still deterministic, not LLM) found additional unique candidates at looser name thresholds — e.g. 377 unique at ≥0.85 and 158 more unique in the 0.80–0.85 band — which is why reconstruction is ongoing work, not a one-shot import.

This is why CiteDiver and CaseDiver care about reporter keys as first-class data, not as optional metadata on a PDF.


Data and market: open corpora, consolidating platforms

Open data got better — and still inherits the citation gap

Harvard's Caselaw Access Project digitized nearly the entire historical print corpus. In March 2024, commercial restrictions ended and ongoing collection shifted toward CourtListener:

CourtListener remains the public front door for free opinions, APIs, and bulk data. Its Citation Lookup and Verification API answers a critical first question: does this citation string resolve to a known case? Free Law Project is also building treatment/citator signal — see Building a Citator with AI, A Progress Report (May 2025).

None of that automatically stamps every scraped Florida opinion with the So. 3d cite Rule 9.800 expects.

The commercial market consolidated sideways into AI

While open data expanded, commercial platforms bought AI-native challengers:

Citators themselves are the oldest moat in the category. Shepard's Citations, begun by Frank Shepard in the nineteenth century, taught the bar that finding a case is worthless without knowing what later courts did to it. Westlaw's KeyCite occupies the same job. Free Law Project's citator work is the open attempt to reclaim it. LawDiver's GoodLaw signal is our product answer. For a fuller comparison of free data versus filing-grade research, see CourtListener vs. LawDiver.


What free data still often lacks

Put the threads together:

  1. Doctrine mostly freed opinion text and (in the Second Circuit) star pagination — not the ongoing editorial machinery of headnotes, digests, and treatment.
  2. Antitrust forced licensing and divestitures around the Thomson/West merger — it did not create universal neutral citations.
  3. Print economics killed series like the Federal Appendix while court rules still speak in reporter language.
  4. Open corpora digitized and harvest opinions at unprecedented scale — yet reporter-key metadata is uneven, especially where courts publish online first and commercial assignment lags or never appears in the free graph.

Florida makes that last point measurable in the charts above. LawDiver maintains a Florida So. 3d reconstruction pipeline for exactly this reason: cite checking and search that claim Florida coverage without reporter keys fail the rules lawyers actually file under.

If you need the practitioner workflow rather than the history, start with How to cite check a brief and Free cite checking software. For cost-side positioning against the incumbent platforms, see Cheap alternatives to Westlaw.


Method and bias disclosure

We build LawDiver and CiteDiver, so we are not neutral. Here is how to check the claims in this article that are ours rather than the historical record.

CourtListener citation-array coverage

The CourtListener coverage figures were produced by querying the public v4 search API — GET /api/rest/v4/search/?q=*&type=o&court=fla&filed_after=YYYY-01-01&filed_before=YYYY-12-31 — paginating with the returned cursor, and counting the fraction of results with a non-empty citation array. For the Florida Supreme Court the sample is the complete annual population in every year except 2010, where it is capped at 1,200. For the District Courts of Appeal, 400 opinions per year were sampled (across fladistctapp1fladistctapp6). A 400-opinion sample returning zero citations is not a sampling artifact: if even 1% of that year's opinions carried a citation, the probability of drawing 400 in a row without one is about 2%. No authentication was used for the original sample; anyone can reproduce this in about twenty lines of Python, and we encourage you to, because if Free Law Project backfills this data tomorrow we would like to be corrected.

Partial re-checks on September 11, 2026 (with API token, still subject to rate limits) reproduced 100% citation attachment for Fla. 2005, 91.4% for Fla. 2015, and 0% for Fla. 2020 and Fla. 2024 (full-year populations).

LawDiver corpus Southern-key rates

The year-by-year Florida tables under "LawDiver corpus" are SQL counts against caselaw_opinion + caselaw_citation in our corpus store (lead opinions, data_quality = 0, Florida court ids). They measure our attachment of Southern reporter keys, including reconstruction — not CourtListener's live API field.

Florida So. 3d reconstruction figures

The Florida So. 3d reconstruction figures describe a dataset in our own repository (volumes 276–432 and the September 9, 2026 assign report) and are not independently verifiable from outside. Treat them as a claim about our pipeline, not as a public fact.

Last verified September 11, 2026. Free Law Project ships quickly and the Third Circuit has not yet ruled in ROSS. If either changes, this article is out of date and we will update it.


Frequently asked questions

Are judicial opinions copyrighted?

The text of U.S. judicial opinions is generally not subject to copyright under the government edicts doctrine and long-standing practice. Publisher-added headnotes, synopses, key numbers, and similar editorial material can be copyrighted. Georgia v. Public.Resource.Org (2020) also held that certain official state code annotations are not copyrightable when they are part of the official law.

Is star pagination copyrighted?

The Eighth Circuit's 1986 Mead decision treated comprehensive West pagination as infringing West's compilation copyright. The Second Circuit's 1998 Matthew Bender star-pagination opinion held that inserting West page numbers into an independent compilation does not infringe. Licensing practice and product design still often treat West-compatible locators as a commercial feature even where doctrine permits more.

Why did the Federal Appendix end?

Thomson Reuters discontinued the Federal Appendix in 2021 after 861 volumes (2001–2021). The series printed non-precedential federal appellate decisions that were already available online; demand for bound volumes collapsed. The opinions remain citable; the F. App'x volume/page address stopped issuing for new decisions.

Does Florida require Southern Reporter citations?

Yes for the ordinary case. Florida Rule of Appellate Procedure 9.800 prescribes So. / So. 2d / So. 3d forms for Florida Supreme Court and DCA opinions published in the Southern Reporter, with Florida Law Weekly and slip-opinion fallbacks when a Southern cite is unavailable.

What is universal or neutral citation?

A vendor- and medium-neutral citation identifies a decision by court, year, and sequential number (often with paragraph pinpoints) instead of commercial volume and page. Some U.S. jurisdictions have adopted variants; many, including Florida for precedential regional-reporter cases, have not replaced commercial reporter cites.

Does CourtListener include reporter citations for every opinion?

No. CourtListener coverage of opinion text is far broader than its attachment of reporter citation metadata for every jurisdiction and year. Florida is a vivid example: many online-first opinions appear without a populated citation array even when a So. 3d cite exists in commercial reporters. Always verify against the API or a citator before assuming a free record is filing-complete.

Is training AI on Westlaw headnotes fair use?

A Delaware federal court rejected ROSS Intelligence's fair-use defense on summary judgment in February 2025 in litigation brought by Thomson Reuters (opinion). That decision is on interlocutory appeal to the Third Circuit (opening brief; answering brief). It is not final. It does not change the public-domain status of opinion text; it concerns publisher-authored editorial content used to build a competing tool.

How is LawDiver involved in this problem?

LawDiver indexes caselaw for search and cite checking. Where free upstream metadata lacks reporter keys that filing rules require — especially Florida So. 3d — we run reconstruction pipelines and expose deterministic citation resolution in CiteDiver. That is product work on top of open data, not a claim that we replace Westlaw or Shepard's wholesale.


Sources

Citation formats

Data and market