ProfText detection rules

Every text you paste into ProfText runs against 30 detection rules. Each rule has a public reference page explaining what it catches, why the pattern matters, and how to fix it. Pages are language-aware (EN / FR / ES / DE / IT where supported).

Run ProfText on your text →

All rules below run in your browser on the free tier. No signup.

AI-generated language patterns (10)

Vocabulary and structural markers that appear in ChatGPT, Claude, Gemini and other LLM output at densities far above natural human writing. Useful both as detection signals and as editing targets for humans aiming for stronger prose.

Em dash (—) — the classic AI writing tell The em dash is the most over-used punctuation mark in ChatGPT and Claude output. Detect overuse before it gives your text away. 'Delve' — the most-flagged AI vocabulary tell 'Delve' and its variants appear in AI output at orders of magnitude above their natural baseline. The single highest-signal vocabulary tell. 'It is crucial / vital / essential' — AI emphasis filler AI models pad arguments with declared importance instead of demonstrating it. Detect filler emphasis phrases that add no information. 'Landscape' — AI corporate cliché 'In the landscape of', 'the evolving landscape', 'navigate the landscape' — AI loves the metaphor and overuses it everywhere. 'Leverage' and 'utilize' — AI corporate jargon 'Leverage', 'utilize' and 'utilise' are over-represented in LLM output versus natural professional writing. Plain 'use' is almost always correct. AI sentence-starter patterns 'Furthermore', 'Moreover', 'In conclusion', 'It is important to note' — AI's predictable connectives at the start of sentences. AI corporate hedging patterns 'It is generally believed', 'many would argue', 'it could be said' — AI hedges to avoid commitment, producing weasel prose. 'Truly', 'really', 'genuinely' — AI empty emphasis Adverbs of emphasis that add nothing. AI models sprinkle these because they soften without committing. 'Was never just for' — AI retroactive reframing pattern 'X was never just for Y — it was always about Z'. A signature AI rhetorical move that retroactively recasts a topic. 'It's worth noting that' — AI introductory filler AI prefaces claims with 'it is worth noting' or 'it is important to mention' rather than just stating them. A dead giveaway.

Citation verification (1)

Detection and verification of academic citations. LLMs hallucinate plausible-looking sources at high rates; ProfText cross-references every detected citation against OpenAlex's open scholarly index.

Academic citation — verify with OpenAlex Detects inline academic citations like '(Smith, 2023)' and verifies them against the OpenAlex database of 250M+ scholarly works.

Temporal vagueness (4)

Time references that age poorly. 'Last year', 'recently', 'in early spring' are all useful in conversation and dangerous in writing meant to last more than a few months.

Relative month without year 'Last May', 'next March', 'this October' — relative month references become ambiguous within twelve months of writing. Relative time period without anchor 'Last week', 'this quarter', 'next fiscal year' — relative periods age out of meaning within months. 'Recently', 'currently', 'nowadays' — vague temporal markers Words like 'recently' and 'currently' carry no anchor date. Readers a year later have no way to know what was meant. 'Early', 'late', 'mid' month or season 'Early spring', 'late October', 'mid-summer' — common shorthand that loses meaning without the year.

Deictic ambiguity (4)

Pointer words like 'above', 'below', 'this', 'the following' depend on layout that may not survive copy-paste, translation, or re-flow into mobile views. Replace with the noun being referred to.

'Above' and 'below' layout references 'As shown above', 'see below' — references that depend on page layout and break in re-flowed or copy-pasted text. 'The following' — vague forward reference 'The following table', 'the following list' — forward pointers that break when content moves or is excerpted. Unanchored 'this' and 'that' 'This is significant' — 'this' what? Anaphoric pronouns without clear referents introduce ambiguity at every paragraph break. References to 'the previous section' 'As discussed in the previous section' — breaks the moment a heading is renamed or content is reordered.

Stale freshness claims (3)

Words like 'latest', 'new', 'up to date' are commitments that decay the day after publication. Use specific dates or version numbers instead.

'Latest' — a commitment that decays the next day 'The latest version', 'our latest research' — claims that are wrong the moment a newer version ships. 'Up to date' — undated freshness claim 'Up-to-date guidance', 'fully up-to-date' — freshness claims with no reference date are unverifiable. 'New' — undated novelty claim 'Our new feature', 'a new approach' — 'new' is a self-discrediting word the moment the feature is six months old.

Vague duration (3)

Phrases like 'a short time', 'a few months', 'from time to time' are impossible to act on. Replace with specific spans.

Vague short durations ('a moment', 'soon', 'shortly') Short-duration shorthand that means anything from seconds to weeks depending on the reader. Vague long durations ('for years', 'over the long term') Long-duration shorthand that obscures specifics the reader needs to act on. 'From' without a clear endpoint 'Available from Monday', 'effective from January' — start dates without explicit duration or end.

Unanchored sequence (2)

'Before', 'after', 'since', 'until' need an explicit reference point. Without one, ordering claims become unverifiable.

'Before' / 'after' without a reference event Sequence words that assume a shared event the reader may not know about. 'Since' / 'until' without a date Open-ended temporal sequencing without anchor dates becomes meaningless as the document ages.

Filler phrases (3)

Boilerplate openers and closers that add length without information. Deleting them rarely removes meaning.

Generic openers ('In today's world', 'In recent times') Boilerplate openers that pad without setting up anything specific. Common in AI drafts and lazy human writing. Generic closers ('In conclusion', 'To sum up') Closers that announce the ending instead of providing one. Strip them — the actual conclusion stands alone. 'In general' — concession without specifics 'In general', 'broadly speaking', 'as a rule' — phrases that hedge claims without identifying the exceptions.