AI-flattened English

Also called: AI voice, homogenised English, generic AI writing

AI-flattened English is the narrow band of register that language models converge on by default — mid-formal, evenly hedged, structurally symmetrical — which reads as competent and carries no information about who wrote it.

It is not bad writing; it is writing with the author removed.

Flattening is a predictable consequence of how the output is chosen. A model asked for a professional email produces something near the centre of professional emails it has seen. Doing that reliably is an achievement. The cost is that the centre is the same place for every user.

The tell is not any single phrase. It is the absence of variance: sentences clustering around one length, every claim hedged to the same degree, paragraphs of near-identical shape. Human writing is lumpier than that, and the lumps are the part readers recognise.

SignalAI-flattened defaultIndividual voice
Sentence length varianceLow — clustered around one lengthHigh — short bursts next to long builds
HedgingUniform across every claimConcentrated where the author is genuinely unsure
Paragraph shapeRepeating three-sentence blocksIrregular, driven by the argument
OpenersRestates the promptEnters where the author habitually enters
Register rangeOne band regardless of readerShifts with audience
Signals that distinguish flattened output from an individual voice

Frequently asked questions

Is AI-flattened English the same as AI-detectable text?
They overlap but are not the same. Detectors look for statistical regularity, which flattening produces, but a heavily edited flat text can pass a detector while still sounding like nobody in particular.
Does asking for a more human tone fix it?
It moves the output to a different average rather than to you. The instruction contains no information about your habits, so it selects another point in the model's distribution.
Do better models flatten less?
Better models follow nuanced instructions more faithfully, which reduces flattening when the instruction actually specifies a voice. With a generic instruction, a better model produces a better-written average.

Read further

Sources and how to cite this page

Cite this page

MyWritingTwin. “AI-flattened English.” Last updated 2026-09-15. https://www.mywritingtwin.com/glossary/ai-flattened-english