To what extent is maternal English affected by these trends? Google AI says that LLMs act as both a sophisticated editing tool and a new agent of linguistic influence that favours specific, AI-typical phrasing. As users increasingly rely on LLMs to polish, summarize, or generate text, these models are shifting human language patterns toward a more formal, standardized, and sometimes “watered-down” form of English.
Key Effects on English Language use are: adoption of “AI-isms”: Human speakers are increasingly adopting words and phrases commonly favored by LLMs, such as delve, underscore, realm, intricate, and meticulous. A study of academic YouTube transcripts found a 35%–51% increase in the usage of these specific terms following the release of ChatGPT.
Reduced Linguistic Diversity: There is a growing risk that human language is becoming less diverse as people rely on AI to generate or refine content. LLMs often produce text that is highly logical but lacks the natural variation, idiomatic richness, and emotional nuance found in human-written English.
“Translationese” and Formulaic Output: When used for translation or text generation, LLMs often produce outputs that feel like “translationese”—a slightly awkward, formulaic style that reflects English-centric patterns in vocabulary and grammar, even when the model is supposedly operating in a different language.
Shift toward Formalism: LLMs tend to favor a “polished,” formal tone, which can homogenize content, making human writing or speech sound more similar to one another.
Impact on Communication and Cognition is Improved Efficiency, Reduced Nuance: While LLMs make text production faster and more accessible, they may reduce the ability to express subtle, context-dependent meanings.
Impact on Critical Thinking: Over-reliance on AI for writing may lead to passive consumption of text, diminishing the active, “metalinguistic” process of crafting language, which is essential for developing an authentic, independent writing voice.
Bidirectional Influence: While early AI research focused on training models to match human language, the trend is shifting toward a bidirectional relationship, where AI models are actively shaping how humans communicate.
English as world second language
Out of the world’s approximately 8 billion inhabitants, 1.4 billion speak English regularly. About 360 million people speak English as their first language: USA; UK, Canada. “71% of Dutch speak good English” or “Netherlands has English Proficiency Index of 652”. Norway (67%), Denmark (69%), Sweden (70%); don’t dub English TV or movies; Finland (65%), Tens of millions in Pakistan and India use English regularly. India 1.4 billion people have English as official, with Hindi; EPI 496. Pakistan English official with Urdu EPI 478; Bangla Desh 476 not official. South Africans have English as one of 11 official languages, but as first language only 9%, compared with Zulu 26%. Nigerian pidgin is understood by up to 75 million “unu dey”, more than UK English. In Japan school leavers have several thousand English words, but not much conversation practice. The 6.5 billion who don’t speak English have poorer access to medicines, university education and tourism. Other contact languages are French, Latin American Spanish, Arabic and Bahasa which give some access to university education. In China only 2% of the 1.4 billion speak good English but they have good access to prestige occupations through Mandarin and are only disadvantaged outside the country. The prestige centres US Harvard, UK Oxford are in decline, but English remains the educated language in the EU despite it not being the language of nay member country!
I suspect the irregularity of English verbs gives the temptation is to use only the present and mark tense adverbially, as Chinese does. I trekked Mont Blanc with a young man from Thailand who was studying for a Master’s in Biochemistry in Sweden. He had a wide English vocabulary but insisted on using only the present tense. Science literature is highly constrained to use past tense passive voice (” these results were found …”), but he managed to work around this. Do ESL teachers detect such a reduction?
Several languages have been honed in the market place, including Chinook, Bahasa and English. Street traders with little formal education often deploy a small English vocabulary to very good effect. For example:
English memsahib she not like summertime Delhi, hot very hot, she go Shimla with train.
The speaker knows that three Hindustani words have been borrowed into English. The subject noun is unmodified from singular but is used stereotypically for plural. Summertime is used adverbially as past tense marker. Both verbs are in the root form, unconjugated but the sense is past habitual.
UK English has many loanwords from Indian in Raj times, but the evidence that Mirpuri and Punjabi backgrounds are dragging English in a novel direction is small. Some Bollywood culture may be known: Yaar, Friend, Achcha, Thik Hain,Okay. … Bas, Chakkar, Dizziness. … Funda, Fundamentals. … Ghanta, Bell (Yeah right).
Prpf Drummond at MMU says that Afro Caribbean English may be seen as “cool” because of music and ganja. Slang imports include bare, rass, mandem or the pragmatic marker “you get me?” Phonological variation include the PRICE vowel: words such as like, might, try rhyme with /cat/, the GOOSE vowel “food, blue, crew” like French tu, and Th-stopping “thing sounds like /ting/.
Emojis
Japanese fretted greatly about the limits of 26 letters and were the first to bring visual imagery into text.
