Themata.AI
Follow coverage
Themata.AI
LLMsAI writing toolsAI detectionContent generation

Report · martinfowler.com

LLM Voice

bliki: LLM Voice

martinfowler.com

Source published October 29, 2026

Summary: 2 min read

🔥🔥🔥🔥🔥

70/100

Summary

Martin Fowler argues that prose with an identifiable “LLM voice” can create an uncanny-valley reaction that reduces readers’ willingness to continue, even when they cannot definitively establish whether a model generated it. He is skeptical of AI-writing detectors because of questions about their accuracy and because readers’ reactions often occur before any detector is consulted. He also cautions that linguistic preferences can reflect generational or regional differences rather than a reliable distinction between human and machine writing. Fowler urges writers not to use LLMs to draft prose or to accept extensive rephrasing suggestions, arguing that human editing often cannot fully remove the model’s voice. He considers AI assistance with outlines potentially acceptable, but says writers should compose the prose themselves and disclose AI writing assistance. Fowler says he uses AI only as a search tool and verifies information against original sources. He views writing as an important mechanism for thinking and comprehension, warning that delegating writing may also delegate parts of those processes. He identifies exceptions, including Jim Highsmith’s use of LLMs to write while living with Parkinson’s disease, translation for non-native English speakers, form letters, and reports intended for automated processes. Fowler says human communication benefits from prose that retains the writer’s humanity.

Key Takeaways

  • Martin Fowler says readers can experience aversion to prose that feels LLM-generated even when they cannot reliably determine its origin.
  • Fowler recommends that authors avoid using LLMs to draft or heavily rephrase prose, while allowing that AI-generated outlines may be useful.
  • Fowler requires contributors to his site to disclose how they used AI for writing; he says he uses AI only for search and verifies information from source material.
  • Fowler argues that writing helps people develop and clarify their own thinking, so delegating writing to LLMs can also reduce independent comprehension.
  • He considers LLM-assisted writing justified in some cases, including disability accommodation, translation, form letters, and documents primarily read by automated systems.

What the discussion said

The thread focused less on whether language models can produce grammatically acceptable prose and more on what readers lose when machine-generated text becomes the default. Many commenters said the practical experience is already bleak: LLM prose feels padded, interchangeable, low on information, and strangely difficult to trust even when no obvious stylistic tell remains. Readers describe skimming for signs of automation before engaging with the argument, then abandoning otherwise promising material because the writing feels trite or directionless. AI-generated fiction flooding community sites was offered as the clearest warning: abundance has made discovery worse, pushing people toward older human-made work. Still, commenters did not reject every writing use. Several saw real value in using models to help people with weaker English communicate technical insight, to critique drafts, or to create impersonal documents where personal voice is irrelevant. The stronger norm was that authors should retain responsibility for meaning and not pass generated prose off as their own work. Some argued writing is itself how people clarify thought and build communicative skill, so outsourcing it risks cognitive atrophy; others countered that tools can widen participation when they amplify a person’s ideas rather than replace them. A few expect models soon to erase detectable cues, while skeptics think human language will adapt to preserve authenticity signals.

Where opinion split

The sharp dispute is whether LLM assistance democratizes expression or hollowes out authorship and reader trust. Supporters argue that polished language should not block valuable technical or personal ideas from people who cannot easily express them. Critics argue that writing is not cosmetic packaging: it demonstrates thought, carries a human voice, and becomes disrespectful when readers must sift machine-generated filler for a real point.

Read original article

Coverage & discussions

Hacker News · martinfowler.com

LLM Voice ↗

Community discussions

Hacker News discussion ↗

Reported comments: 263 · 15 saved in sample

Analysis ready

Community Sentiment

Negative

Positives

  • LLMs can let people with limited writing fluency convey technical insight and personal stories that polished-English gatekeeping would otherwise bury.
  • Using models as critics, brainstorming partners, or drafting help can preserve an author's judgment while reducing the friction of routine, impersonal documents.
  • For formal writing with an intentionally standardized voice, such as briefs or design documents, AI assistance can be a sensible efficiency tool rather than an artistic betrayal.

Concerns

  • Machine-written prose is widely experienced as verbose, generic, and low-density, turning promising articles into work readers abandon before reaching the useful idea.
  • Readers increasingly inspect text for AI signals before evaluating its argument, a trust tax that makes online writing feel less like communication and more like filtering spam.
  • Letting models do both writing and comprehension risks outsourcing the mental work through which authors develop ideas, arguments, and a distinctive voice.
  • AI content floods, especially in fiction, make discovery harder because near-limitless mediocre output crowds out work readers believe came from an actual person.
  • Constant exposure to generated prose may train humans into its flattened style, spreading the same voice even where no model was used.

Related Articles

On the non-use of AI in my writing process

Charlie Stross – On the non-use of AI in my writing process

Aug 1, 2026

The Future of Everything is Lies, I Guess: Where Do We Go From Here?

The future of everything is lies, I guess: Where do we go from here?

Apr 16, 2026

The Problem With LLMs

The Problem with LLMs

Feb 12, 2026

The LLM Critics Are Right. I Use LLMs Anyway.

The LLM Critics Are Right. I Use LLMs Anyway

Jul 16, 2026

Detecting LLM-Generated Web Fiction with "Classical" Machine Learning (AIGC Text Detection)

Detecting LLM-Generated Texts with “Classical” Machine Learning

Jul 16, 2026