The list of "tell"words for LLM speak grows, but these Bots are now teaching us to speak like them ..
According to an article in the Washington Post on 20 August 2025 by Adam Aleksic "people are starting to talk like ChatGTP". The previous post here on the Realm pointed to some words that have been nominated as indicators that a Large Language Model (LLM) is the likely author of the piece. According to Mr Aleksic's article we can now add to our list of"delve", "adept", "realm" and "meticulous", the words "intricate" and "commendable".
As he says elsewhere, it is becoming hard to keep up. But his more telling point is that as LLMs become more and more prevalent contributors in our discourse, it's not just the bots that are using these words, we've all started using them more too; in our own speech.
It seems that the feedback loop from the LLMs gulping up so much of their own outputs when scraping content from the web, is intensifying this linguistic shift. Aleksic says:
... In the two years since ChatGPT launched in late 2022, the appearance of “delve” in academic publishing saw a tenfold increase as researchers began turning to AI for help with their papers. ..
He then says:
...but so many words are appearing with unnatural frequency that we can’t possibly avoid them all. Are we also supposed to stop using the chatbot-overused “inquiry”? Or “surpass”?
Note to self: there's two more words to add to our list of LLM speak words to eschew (that now sounds like a suspicious word too) , to avoid being mistaken for a bot. My English might have to start being more Anglo-Saxon soon, to reduce this risk. Shit.
And apparently these LLMs aren't neutral. They have embedded biases. Be damned! It seems they have the ability to change the way we think... Maybe we should all start speaking Mandarin or Cheyenne ...?
Comments