Every practitioner has done it. The deadline is tomorrow, there are nine authorities to get through, and the headnote says exactly what you need it to say. So the proposition goes into the note, the citation goes next to it, and the judgment itself stays unopened. Most of the time nothing happens, which is precisely why the habit survives. When something does happen, it happens in open court, and it happens in a form that is difficult to recover from: the bench turns to the paragraph and the paragraph does not say what your submission says it says.
The problem is not that headnotes are badly made. Many are made with great care by editors who read the judgment far more closely than the average advocate under deadline. The problem is categorical. A headnote is a description of the law. It is not the law. It is written by a person who is not the court, it is not part of the judgment, and it has no authority of its own. Once that is properly internalised, the professional line falls into place: cite only what you have verified in the text.
The line worth holding
A headnote is enough to decide whether to open the judgment. It is not enough to decide anything else. Every other use, citing, quoting, advising, submitting, requires that you have read the passage in its own context.
What a Headnote Actually Is
A law report is an edited product. The court delivers a judgment; a reporter selects it as worth reporting, and an editorial team prepares the apparatus around the text: the catchwords, the headnote, the summary of counsel's arguments, the list of cases referred to, and the marginal letters or paragraph markers that let a reader pinpoint a passage. The judgment sits inside all of that, and only the judgment came from the court.
The headnote's function is navigational. It exists so that a reader flipping through a volume can tell within seconds whether this case is about the thing they are researching. That is a genuinely valuable job, done well. It is not the same job as stating a holding with the precision required to build an argument on it, and it was never designed to be.
There is a simple structural proof that a headnote is an interpretation rather than a transcription. Different reporters headnote the same judgment differently. Two competent editorial teams reading one judgment produce summaries of different length, different emphasis, and sometimes different propositions, because summarising requires choosing what matters and choices differ. If a headnote were merely a restatement of what the court said, the versions would converge. They do not, and that divergence is the whole argument in miniature.
The moment two reporters summarise one judgment in two ways, the question is settled: a headnote is a reading of the judgment, and readings can be wrong.
The Failure Modes, Structurally
It is more useful to name the ways summaries fail than to hunt for anecdotes. These failures are inherent in compression, and they recur across every reporter and every jurisdiction.
Compression drops the qualifications
Judgments are full of conditions. The court holds X, provided the record shows Y, in a case where the parties had not raised Z, and subject to what is said later about W. A headnote that carried all of that would be as long as the judgment. So the qualifications go, and what remains is a clean sentence that is broader than anything the court actually decided. The reader inherits the breadth and none of the conditions, and cites the proposition in a case where the conditions are absent.
The facts that made the proposition true disappear
A holding is a holding about something. Strip the facts and you are left with an abstraction that looks portable and is not. This is where distinguishing happens: the opponent's answer is almost never that your proposition is wrong, it is that the case that produced it was a different case. A researcher who has read only the headnote has no answer to that, because the facts that would supply the answer were the first thing the summary removed.
Ratio and obiter are flattened into one
A judgment contains a great deal that is not the ratio: observations made in passing, responses to arguments not necessary for the decision, general remarks about the state of the law. A headnote presents propositions in a uniform register. It rarely marks which sentence carried the decision and which was said on the way past. So a considered observation and the actual holding arrive on the page looking identical, and the reader has no way of telling them apart without the text.
A multi issue judgment becomes a list
Long judgments decide several questions, and the answers interact: the court may reach the second question only because of how it answered the first, or may decide a point conditionally. Turned into a bulleted set of propositions, those dependencies vanish. Each item reads as free standing when in the judgment none of them is.
The editor's view of the case is embedded in it
A headnote states what the editor understood the case to decide, at the time the report was prepared. Later courts may read the same judgment differently, narrowing it, confining it to its facts, or treating a passage as the ratio that the original headnote treated as incidental. The headnote is frozen at the date of the report. The judgment's meaning, in practice, is not.
The specific error that a bench notices
Quoting a headnote in a pleading or a written submission as though it were the language of the court. It is visible immediately: the words are not in the judgment, and when the bench turns to the paragraph the discrepancy is on the record. It costs more than the point. It costs the bench's willingness to take the rest of your citations at face value.
Catchwords, Marginal Letters and the Rest of the Apparatus
The other editorial furniture deserves the same treatment. Catchwords, the string of subject terms at the head of a report, exist so the case can be indexed and found. They are a filing system: they tell you which drawer the case was put in, not what it holds, and they are chosen with an eye to the index rather than to your question.
The letters running down the margin of a printed report, and the placitum markers some series use, are pinpointing devices belonging to that edition. They do not travel to another edition or to the court's own text. Where the court has numbered its own paragraphs, that numbering belongs to the judgment and travels everywhere, which is why a paragraph pinpoint is worth more than a page and letter pinpoint to a reader holding something else.
The summary of counsel's arguments, where a report carries one, is regularly mistaken for the court's reasoning. It is a record of what was submitted, and a submission that was made and rejected reads, out of context, exactly like one that was made and accepted.
AI Summaries Are the Same Artefact, With One New Failure
An AI generated case summary belongs to the same category as a headnote. It is a compression of a judgment produced by something that is not the court, carrying an interpretation, and inheriting every limitation described above: dropped qualifications, missing facts, ratio and obiter flattened, dependencies lost. Nothing about the technology changes the category. The old rule applies unchanged.
There is, however, one failure mode that is more pronounced. A human headnote written under editorial convention tends to signal its own hedging: it uses the language of the report, it often quotes, and where the editor is unsure the prose becomes noticeably careful. A generated summary produces fluent, confident prose regardless of how well grounded it is. The characteristic error is a qualifier that shifts. The judgment says a court may ordinarily decline; the summary says a court will decline. The judgment says the position is different where the record shows something; the summary drops the condition. Each is a small movement, each reads perfectly, and each changes the proposition you would cite it for.
The practical rule is therefore identical and slightly stricter. Use a generated summary to triage, exactly as you would use a headnote. Never let a sentence from one enter a document you file. And verify every citation attached to a generated summary against the judgment itself, because an accurate summary attached to the wrong case is a failure that a fluent paraphrase does nothing to reveal.
What a Headnote Can and Cannot Tell You
| The question you have | Can a headnote answer it | Why |
|---|---|---|
| Is this case about my subject at all | Yes, reliably | This is the job the headnote was built for. Catchwords and the opening lines are designed for exactly this triage. |
| Roughly which way did the court go | Usually | The direction of the outcome is hard to compress wrongly, though the reasons for it are easy to lose. |
| What exactly did the court hold | No | The holding is a proposition tied to facts and conditions. Compression removes precisely the tying. |
| Is this passage ratio or obiter | No | Headnotes present propositions in one register and rarely mark which carried the decision. |
| Can this be distinguished on my facts | No | The facts the court treated as material are the first casualty of summarising, and they are the whole of a distinguishing argument. |
| What are the court's exact words | No | The headnote is written by an editor. Its words are not the court's words, and quoting them as such is visible. |
| Has a later court read this case differently | No | The headnote is frozen at the date of the report. Subsequent treatment sits outside it entirely. |
| Was there a separate or dissenting opinion | Sometimes, in outline | Reports vary. Where a dissent is noted at all it is usually noted briefly, and its reasoning is rarely summarised in usable form. |
A Triage Rule You Can Actually Run
Nobody reads every judgment in full, and pretending otherwise produces advice that gets ignored. What works is a rule that is explicit about which uses require a full read, so the decision is made once rather than negotiated with yourself under deadline. It applies identically to all three of the summaries you are likely to be holding.
The printed headnote
Prepared by an editorial team, frozen at the date of the report, and different from the next reporter's version of the same judgment. Careful work, and still an editor's reading rather than the court's words.
The database summary
Often a headnote by another name, sometimes an abridgement of one, sometimes assembled to fit a template. Its provenance is usually not stated on the screen, which makes it harder to weigh than a printed headnote, not easier.
The generated summary
The same category of artefact, produced faster and read more confidently. It carries every limitation of the other two, plus a tendency to smooth a qualifier out of a proposition without any visible seam.
The rule, in one line
A summary is sufficient for deciding whether to open the judgment. A full read is non negotiable if you will cite it, quote it, rely on it in a written submission, advise a client on it, or if the point is close. There is no fourth category.
Triage the field on summaries, quickly and without guilt
Run your search, read the headnotes or summaries of everything returned, and sort into three piles: plainly irrelevant, possibly relevant, and probably central. This is the correct use of a summary and it should be fast. Do not take notes from the summaries at this stage, because notes taken here have a way of surviving into the final memo.
Open everything in the central pile, at source
Go to the judgment text itself rather than to a reproduction of it. Where the court's own record is available, use it. This is the point at which the research stops being a list of results and starts being research.
Read the facts the court treated as material
Not the full narrative, but what the court itself identified as the operative facts, usually in its own framing of the case. These are what make the proposition portable or not, and they are what your opponent will use to distinguish the case. If you cannot state them in two sentences, you have not read enough.
Find the questions the court framed for itself
Most reasoned judgments state the issues, either formally or in the course of the reasoning. The framing tells you the scope of the decision. A proposition answered under issue three is authority for issue three, whatever a summary has done with it.
Read the operative reasoning, and locate your paragraph inside it
Find the passage you want, then read what comes before and after it. A sentence that survives on its own may be the court reciting a submission, restating an authority it is about to reject, or setting up a contrast. Context is what distinguishes a holding from a hypothetical, and it is invisible in an extract.
Read the final order
The operative portion tells you what the court actually did, which is the surest indicator of what it actually decided. Reasoning that points one way and an order that goes the other way is a signal to read the whole thing again more carefully.
Check for a separate or dissenting opinion
A concurring opinion may narrow the majority's reasoning, and a dissent tells you where the fault line is and what the strongest contrary argument looks like. Summaries handle separate opinions poorly. Read at least enough of them to know what they say.
Record the paragraph number and the exact words you will rely on
Copy the sentence, with its paragraph number, into your note as you read. This single habit is what prevents the headnote proposition from creeping back in later, and it means the appendix of your memo is built as you go rather than reconstructed the night before.
A short checklist before a citation leaves your desk
- Have I opened the judgment itself, not a summary, a database extract, or another document quoting it
- Is the sentence I am relying on in the judgment, in those words, at the paragraph I have noted
- Is that sentence part of the reasoning that decided the case, or an observation made in passing
- Do I know the facts the court treated as material, well enough to answer a distinguishing argument
- Does the final order match the proposition I am attributing to the case
- Have I read enough of any separate opinion to know it does not undercut me
- Have I checked what has happened to this judgment since, rather than assuming a clean report means a clean authority
Where a Research Platform Helps, and Where It Does Not
A research platform is genuinely useful at the triage stage and genuinely dangerous if it is allowed past it. CourtMesh runs one search across the Supreme Court, all 25 High Courts, District Courts and Tribunals, sourced from official government portals, with roughly 310 million records keyword-searchable, and every result links to the judgment text. That last point is the one that matters for this article: the endpoint of a search should always be the document, not a description of it.
AI case analysis and related-case discovery are available across an analysed subset, roughly 2 million records semantically searchable, with citation relationships typed as followed, distinguished, overruled and referred. Used properly, that shortens the triage and tells you which of the fifty results are worth opening. It is a sorting instrument.
The temptation the tool creates
A generated summary is more fluent than a printed headnote and arrives faster, which makes it easier to rely on and no more reliable. Every limitation in this article applies to it, and the shifting qualifier applies to it more. Treat any summary a machine produces the way you would treat a headnote written by a stranger under time pressure: useful for deciding what to open, and never a source you cite.
The official record prevails
Whatever a research tool displays, the record of the issuing court is the authoritative one. Confirm the citation, the date, the bench and the paragraph you rely on against that record before anything goes into a filing.
Read the case you are citing
The professional line has not moved. Summaries, printed or generated, are navigation aids, and the only thing you may safely cite is what you have read in the judgment. CourtMesh is built to make that read short rather than to replace it: one search across the Supreme Court, all 25 High Courts, District Courts and Tribunals from official government portals, structured filters for court, year, judge, case type and date range, and every result linked to the judgment text. Triage on the summary. Cite from the paragraph.
Explore CourtMesh
