TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
A preprint by three linguists compares text from 14 language models with human writing and identifies six language patterns the authors say can make AI text harder to process. The study has not been peer reviewed and did not test whether readers actually understood the text less well; its practical advice is to edit for clarity rather than rely on AI detectors.
Three linguists have identified six language patterns that appeared more often in writing from advanced AI models than in comparable human text, in a preprint posted on October 21, 2026. The researchers say the patterns may make text harder to process, though they did not test whether readers understood it less well.
The paper, “Why is AI-generated language so hard to understand?”, compares output from 14 language models across casual chats, coding sessions and short persuasive essays. The authors are Thomas Hikaru Clark of New York University and Matias B. Martin and Kyle Mahowald of the University of Texas at Austin. Their comparison used human writing from sources such as film and television subtitles, StackExchange, student essays and Wikipedia.
The six patterns are rare word combinations, long gaps between words that belong together, omitted “that” clauses, strings of three or more nouns, nouns used as verbs and descriptions that give inanimate objects human actions. All nine frontier or near-frontier models in the study used more rare word combinations and more descriptions of objects acting like living things than the human comparison texts did. The authors also report that these patterns increased across base, instruction-tuned and reasoning versions in the Olmo model family.
To test whether models had chosen an unusual phrasing when a plainer alternative was available, the researchers translated sentences into Japanese and back into English. They treated a pattern becoming rarer after this round trip as evidence that the model had selected it despite the availability of more ordinary wording. The paper’s appendix puts the data collection’s API cost at $251.19.
1Rare word combinations
Word pairs that people seldom put side by side.
Before (illustration)Your onboarding email has a trust leak.
After (illustration)Your first email to new customers makes them doubt you.
Fix: Use the words that your reader would say out loud.
2Long gaps
The subject waits a long time for its verb.
Before (illustration)The report the Berlin team, after three rounds with legal, sent on Friday was wrong.
After (illustration)The Berlin team sent the report on Friday, after three rounds with legal. It was wrong.
Fix: Verb right after its subject. Split long sentences.
3The missing “that”
One word saved, one stumble caused.
Before (illustration)The fix the vendor shipped broke the export.
After (illustration)The vendor shipped a fix, and that fix broke the export.
Fix: Keep “that” when it shows how the parts fit.
4Noun stacks
Three or more nouns in a row.
Before (illustration)Please review the customer data retention policy update draft.
After (illustration)Please review the draft update to our rules on how long we keep customer data.
Fix: Unpack with “of”, “for” or “to”, or cut words.
5Nouns used as verbs
Ad-hoc conversions like “calendar it”.
Before (illustration)Let’s roadmap this and calendar the review.
After (illustration)Let’s add this to the roadmap and put the review in the calendar.
Fix: Use the plain verb, even if it takes two more words.
6Objects that act alive
Code “swallows” an error, a problem “bites”.
Before (illustration)This setting quietly kills your conversion rate.
After (illustration)With this setting, fewer visitors buy.
Fix: Name the person or the cause, then say what happens.
Editing Can Reduce Reading Friction
The findings give writers and editors a practical checklist for revising AI-assisted text: replace unusual word pairings with familiar language, bring related words closer together and break up dense noun strings. The paper’s examples suggest that small sentence-level edits may make prose easier to follow without changing its core meaning.
The authors argue that training for tasks such as mathematics, coding and reasoning may reward dense answers more than clear ones. That is their proposed explanation, not a demonstrated cause. The study also does not establish that every instance of these patterns confuses readers, or that all AI-generated writing is difficult. Its contribution is to identify recurring differences that can guide editing and further research.
The paper cautions against treating automated detection as the remedy. The source article says AI detectors have a weak record and have mistakenly flagged human writers. Editing for clarity addresses the text itself; a detector’s label alone cannot establish how a passage was written.
Products Worth Considering
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
Apple iPad Air (4th Generation, 10.9-inch, Wi-Fi, 256GB) – Space Gray
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How Researchers Compared the Text
The researchers gathered text in three settings: 120 casual chats with a simulated user, 45 coding sessions on real GitHub projects and 980 short persuasive essays. The model set included nine frontier or near-frontier systems and five smaller open models from the Llama and Olmo families. Human comparisons were selected to match the broad type of writing being studied.
The paper says smaller models with lighter post-training wrote more like people on some measures. It also reports that Claude Opus 5.5 scored lower than Opus 5 on 19 of 21 difficulty comparisons. These are results within the study’s measures and sample; they do not establish a general ranking of model quality or readability for every task.
The six patterns are not presented as a detector that can identify AI authorship. They are features the authors counted when comparing samples. The source article’s before-and-after editing examples are its own illustrations, rather than model quotations; the quoted Claude sentence comes from the paper’s Table 2.
“Why is AI-generated language so hard to understand?”
— The researchers, as quoted in the paper’s title
Reader Comprehension Was Not Tested
The work is an arXiv preprint and has not been peer reviewed, according to the source material. It focuses on English and does not report experiments measuring whether real readers understood the sampled text more slowly or less accurately. The link between the six patterns and reading difficulty draws on earlier studies, but this paper does not independently confirm that effect in its AI samples.
The source does not provide enough detail to assess how representative the simulated chats, coding sessions, essays or human comparison sets are of everyday AI use. It also remains unclear how well the findings apply to other languages, tasks and models beyond those examined.
Further Tests Could Check Readability
The next step for researchers would be to test whether readers actually find passages containing these patterns harder to understand, and whether edits improve comprehension. Peer review and replication with other models, writing tasks and languages could also clarify how broadly the findings apply.
For readers editing AI-assisted work, the source article recommends revising sentences directly: choose familiar word combinations, move related words closer together and simplify dense noun constructions. The study does not establish a single editing method as effective for every passage, so clarity still needs to be judged in context.
Key Questions
What did the study find?
The preprint identifies six language patterns that appeared more often in AI text than in comparable human writing. It does not establish that every example of those patterns is hard to understand.
Has the study been peer reviewed?
No. The source describes it as an arXiv preprint posted October 21, 2026, and says it has not been peer reviewed.
Did researchers test whether people understood the AI text?
No. The study counted language patterns, but the authors did not test whether real readers understood the text less well or more slowly.
What can writers do to make AI-assisted text clearer?
The source recommends replacing unusual word pairings with familiar wording, keeping related words close together and simplifying long noun strings. These are editing suggestions, not a guarantee that every passage will become easier to read.
Source: Thorsten Meyer AI
