Quibberfluff. Snizzlewump. Wobblefizz. Three pseudo-words coined by ChatGPT last week. Heddwen Newton, an English teacher, had challenged the bot to “generate a list of nonsense words”. Sure, said the software, conjuring blibberquark to zonkledoodle.

Logically, if there’s any logic to be found in nonsense, the gibberish tells us several things. First, the algorithm is fond of compounds and polysyllables, much in the jabberwocky vein. Second, the guff is pronounceable, with a seemly ratio of vowels to consonants, including the orthodox QU pairing. As nonsense goes, quackleflop makes sense, though Heddwen had her qualms.

The issue lay with lrtsjerk, which didn’t appear on the list. The curio emerged last month, an odd fish snagged by Google Alerts. As a fan of English, and especially its evolution, Newton writes a monthly newsletter called English in Progress, a vibrant link-fest to new slang, jargon, accents and everything else. By fluke, lrtsjerk had crossed her path due to its alleged definition: “a term coined to describe a linguistic trend characterised by the intentional misspelling or alteration of words.”

Says whom, asked the language blogger. More probing revealed more meanings, each one zanier than the last. According to a series of iffy websites, lrtsjerk meant a group of online jerks, a magical land, a radical means of thinking or “a full-bodied workout with ancient roots”. Compared to ChatGPT’s nonsense words, this glossary was garbage.

Yet there it was, in plain sight. Whole screeds supporting a word no dictionary or news source had ever encountered. Diligent sleuthing dated the weirdness back to September 2023, perhaps the first citation. It’s hard to be sure. Far clearer was the invisible hand of AI-generated text at play, seeking to dignify the novelty.

As Heddwen explains, “AI text stands out, as it’s often long-winded, repetitive, contradictory and doesn’t cite sources.” Other tells are also evident, for now at least, including the layout of preliminary contents, plus Q&A on the bottom. But that may change, making this lrtsjerk business an omen, and not a good one. For a non-word to be granted bogus word-status, across six hokey sites at least, may chill a lexicographer’s blood.

Loading

Covfefe, of course, the Trump gaffe of 2017, was mumbo-jumbo of human origin. Born of accident or one-off circumstances, the nonce word denoted nothing until it came to mean the error itself, a droll shorthand for a fat-finger error. Could lrtsjerk lay similar claims, or is it more insidious? Less a typo than an “algo”, born of bytes or sly design? Is it the first AI-born word in our lifetime, or the prank of a nameless interloper?

Or maybe a ghost word, a bogus term that starts out as fabricated, only to be listed as genuine by a reference work due to lexical misinterpretation? Dord, say, bobbed up in Websters 1934 edition as a synonym of density. Wrong. The non-word was a misread of ‘D or d’, standing for density, yet the lapse became briefly legit.