FILE: ARCH-RPT-2026 / METHOD / STATUS: PUBLISHED
Method: what was counted and what was not
This section is the report's spine. Nothing in it may be softened in editing.
Read this page after the report, or when you need to check a number. Each section states one thing that was counted.
WHAT THE CORPUS IS.
3,174 phrasings, recorded across three mining passes in July 2026. They are held in three version-controlled documents in The Archivist Method's repository. Each phrasing is a plain sentence a person types or says about a behavior they have caught themselves repeating. It is recorded before they know the pattern has a name.
WHAT THE CORPUS IS NOT.
Nobody measured how many people typed anything. No number in this report counts searches, people, frequencies, or trends. A sentence of the form "N people search for X" cannot be built from this data, and this report does not contain one. Where this report says 141 or 43 or 268, it is counting recorded phrasings. Nothing else.
THE TWO EVIDENCE GRADES, KEPT SEPARATE.
Every quoted phrasing in this report carries one of two labels. The difference is never blurred:
- Typed title. A real public title a real person typed, recorded word for word or nearly so in the corpus's grounding tables. There are 80 of these: 22 confirmed in the second mining pass, 58 in the third. The first pass confirmed zero raw titles and says so in its own text.
- Corpus phrasing. Built from those confirmed anchors by a written register method, in the vocabulary the anchors establish. The corpus's own instruction is followed here: treat the vocabulary as confirmed and the volume as a working guess, not a fact.
HOW THE NUMBERS WERE VERIFIED.
Every figure is labeled at source in one of two ways. STATED means the mining document's own count, with a file-and-line receipt. RECOUNTED means a script reproduced it in August 2026. The third mining pass reproduces exactly: all eighteen of its cells and all three of its quarantine mark counts. The corpus's canonical total is 3,174, as the mining documents state it. The floor a script can reproduce on its own is 2,837 register-line phrasings. Two findings in this report rest on that floor, the scenes-and-labels split and the vault, and both say so.
THREE DISCREPANCIES, STATED OPENLY.
The recount found three gaps between stated and mechanical numbers. All three are recorded here and none is silent:
- The second mining pass states 28 diagnosis-naming phrasings. The mechanical count is 34.
- The second pass's six observer files each state one phrasing more than their mechanical register count: 135 stated against 129 counted.
- The first pass states 310 territory phrasings. The mechanical count of register positions is 412. The difference fits a de-duplication against the pattern files that was never written down.
None of the three touches a finding in this report. Every finding above rests on an exactly verified number or on the mechanical floor.
THE CLASSIFICATION BEHIND FINDING ONE.
Scenes and labels are sorted by mining register, which is the corpus's own unit of organization. They are not sorted phrasing by phrasing. The scene registers are situation, body, night, paradox, aftermath and crisis-adjacent. The label registers are vernacular, misfiled label and diagnostic. The behavior questions, meaning cause, action and spoken, count as neither and are shown in the receipt. Phrasings inside a register vary. The ratio is exact about registers, not about every phrasing.
THE EXCLUSIONS.
Three kinds of phrasing appear nowhere in this report, in any form: anything marked as typed after the event, anything typed in present distress, and anything naming a diagnosis. The corpus itself records that quarantine. Whole territories are excluded for good by standing ruling, including symptom-checking, substance territory, and compensatory exercise. Active self-harm territory is absent from the corpus itself, by design: the miners refused to record it. A person searching those words needs a crisis service in the first result, not a pattern file.
THE 1AM STANDARD.
Every "what it means" passage in this report is written to one bar. A stressed, non-technical reader can read it at one in the morning. No term is left standing that the reader would have to look up. No comfort is offered that is not also accurate.
QUOTING LAW.
Quoted searcher phrasings may contain vocabulary The Archivist Method never uses in its own voice. They appear under a scoped, founder-ruled extension of the archive's quoted-searcher exemption, recorded in the repository's language law on 2026-08-22. The rule is simple. The report quotes the searcher and answers in the archive's own vocabulary. It never quotes a diagnosis. It never quotes the vault.
PROVENANCE, VERSION-CONTROLLED.
This report derives from docs/report/REPORT-2026-ANALYSIS.md at repository commit d8ae1c43. That analysis cites the three corpus documents by path and line for every figure: docs/seo/PATTERN-PAGE-CLUSTERS.md, docs/seo/PATTERN-PAGE-CLUSTERS-HARVEST-2.md, and docs/seo/PATTERN-PAGE-CLUSTERS-HARVEST-3.md. The analysis, the rulings, and this report are all in version control. Every number in this report can be walked back to a file, a line, and a date.
THE REPORT
The findings this method stands under are the report itself. The 1AM Search Report, 2026 edition
HOW TO CITE THIS REPORT
https://thearchivistmethod.com/report/2026
SHORT FORM, FOR RUNNING TEXT
According to The Archivist Method's 2026 report, [finding]. The 1AM Search Report, 2026 edition.
REFERENCE FORM
The Archivist Method (2026). The 1AM Search Report, 2026 edition. The Archivist Method. https://thearchivistmethod.com/report/2026
FOR JOURNALISTS, ONE LINE
Source: The 1AM Search Report (2026), The Archivist Method's annual filing on the phrasings people type about their own repeating behavior; figures are counts of recorded phrasings, not of searches or people.
Attribution is required in the branded form above. The report may be quoted freely with attribution; the vault may not be quoted by anyone, because it is not published.