Get in Touch

Phrase study · 18 measured phrases

Eighteen phrases we can honestly number

See the eighteen three-word phrases that survived measurement, with document counts and interval estimates.

On this page
  1. The finding
  2. The table that nearly shipped
  3. Why the table is so small
  4. Does it replicate?
  5. The eighteen rows
  6. The artefact that got through
  7. What travels with the panel
Published
31 August 2026
Measured
30 August 2026, re-verified at source 31 August 2026
Corpus
The 5,558-document long-form corpus: 922 AI documents across 13 models, 4,636 human documents, split into mining and held-out halves by id hash
Detector involved
None — a corpus lookup, which is exactly why it can be checked

01 / The finding

Honest phrase evidence is small, interval-shaped, and one judgement call from wrong

The first table's strongest "AI phrase"per cent of

30.1× on the mining half — because our AI corpus was generated in British English and the human half is largely American. 29.9% of AI documents contain "per cent" against 1.4% of human ones. A spelling detector wearing an AI detector's label, caught before shipping and published here.

Phrases that clear the count floor, whole corpus2,503 / 251 / 21

Three-, four- and five-word phrases appearing in at least five documents on both sides. That is why nothing longer than three words is published: 21 five-word phrases is not a table. A commercial panel shows longer phrases because it indexes millions of documents; ours holds thousands and says so.

Held-out replication of the top fifty0/50 below 1×

Ranked on the mining half, reported on the held-out half: log-ratio correlation 0.73, median held-out ratio 10.3×, and not one phrase reversed. The effect is real. What it is not is precise — intervals commonly span a factor of five.

02 / The near miss

The first table was a British-English detector, and it is recorded because it nearly shipped

Built on raw text, the three strongest "AI phrases" in the corpus were "per cent of" (30.1×), "per cent in" (28.3×) and "cent of the" (28.0×) — the same artefact three times. Our AI half writes "per cent" and "-our" spellings (47.2% of AI documents against 11.6% of human); our human half is largely American. The fix normalises spelling to one convention on both sides before anything is counted, and the "per cent" family disappears from the ranking entirely. The same failure is visible in the competitor's own published panel, where a domain phrase from the user's document is ranked at 307× — quotable, specific, and measuring the topic rather than the author.

03 / The rules

Document frequency, a count floor on both sides, and a register control

Measured at a rule that no longer ships

Corpus
Mining half 467 AI / 2,315 human; held-out half 455 AI / 2,321 human, assigned by id hash. Phrases are ranked and register-controlled on the mining half and REPORTED on the held-out half — selecting on the numbers you then publish is how a table ends up describing its own noise.
Retired flag point
0.9855 / 0.9763
Detector
tier3-cycle2-e5small-fp32.onnx, SHA-256 e313ab00de1fffd2…4d2788d (cycle 2, superseded 1 September 2026)
Runtime
Pure counting after spelling normalisation. Document frequency, not raw frequency: a phrase used eleven times in one verbose white paper is one document's worth of evidence.
Measured
30 August 2026
Also
A phrase must appear in at least five documents on each side of the mining half, must lean AI inside every register with enough documents to test (19 of the top 40 do; the 21 that fail do so for want of power, none reversed), and its held-out interval must exclude 1.

The rule that ships today is margin 3.570935 / gap 0.34 (display 0.9679 / 0.9562). The figures under this stamp answer the same question at a flag point this tool no longer uses, and they are not a description of the tool as it runs now.

04 / Replication

The finding under the table: machine prose argues by explicit contrast

Most of the surviving phrases are contrastive or hedging frames — "rather than a", "is not simply", "not the same", "the gap between", "the limits of". Set beside the repetition finding — machine writing shares only 2.1% of content words between neighbouring sentences against 6.3% for human prose — the two are one description of how the prose is built: machine prose argues by explicit contrast where people repeat and accumulate. The house writing guidelines banned "it's not just X, it's Y" on taste; 922 documents arrived at the same judgement by counting.

05 / The table

The eighteen rows, exactly as the checker serves them

This table is read from the same JSON file the checker fetches — phrase-ratios-v1, measured 2026-08-30 — not retyped. Every ratio is the held-out 95% interval; every row carries its document counts, and the weakest rows sit on nine or ten documents, visibly.

The shipped phrase table with held-out intervals and counts
PhraseHeld-out ratio, 95% intervalRegisters testedAI docsHuman docs
the literature is7.1–55.5×3174
the evidence base19.9–174.8×4403
the result is10.6–65.4×4285
rather than an5.2–19.1×32613
not the same7.8–50.5×6215
but as a6–35.2×5186
is not simply11.9–161.6×3212
rather than as10.2–75.4×4244
question of whether4.1–14.5×32315
the gap between5.2–18.1×72714
a fraction of2.6–15.5×3108
a mixture of3.5–23.4×4116
evidence from the3.9–29.3×3115
the first is3.9–13.5×52316
rather than a8.5–17×710042
should therefore be5.6–59.2×3123
the limits of5.1–23.6×3209
ways that are6.2–49.9×3154

06 / The one that got through

"the bank of" passed every automatic filter, and was cut by judgement

One phrase — "the bank of", 2.5–16.7× held out, on 9 AI and 7 human documents — cleared the count floor, the register control and the held-out interval test, and is still not a machine tell. Our AI half writes about banking across several registers, so a control that compares a phrase's lean between registers cannot separate subject from style when the subject is everywhere. It was excluded on credibility, not statistics: a reader pasting a banking article who sees their own subject marked as an AI phrase will conclude the tool is not thinking, and on that row they would be right.

The exclusion is not a quiet edit. It lives in the table-building script as a named entry with its reason, and is emitted into the shipped JSON under excluded_by_judgement, so it is reproducible and cannot be mistaken for tuning the table until it read well. What it means for the other rows is also stated: at least one topical artefact reached the final filter, others may sit below the point where they were obvious, and the counts are printed against every row so a reader can see exactly how thin each one is.

07 / Limits

What travels with the panel, wherever it renders

A phrase in your draft is not evidence about your draft. These are tendencies across thousands of documents; people write every one of them, and the panel leads with that sentence before any number.

Small corpus, wide intervals, long-form only. 922 AI documents against a commercial service's millions is why the table stops at three words and eighteen rows. Short marketing, SEO and social copy are not represented and cannot be — every sample this programme owns of those registers sits inside the training set. 268 of the 922 AI documents touch a cycle-2 split; no classifier is involved in a phrase count, but the documents are not guaranteed independent of the model either, and the caveat ships.

Apply the method

Paste a draft and see which of the eighteen appear — counts attached.