AI’s Hallucination Hall of Fame: How Google’s Bot Keeps Making Up My Wives

By Martin Rowson

When I first stumbled upon Google’s AI Overview feature, I thought I’d found the perfect tool for procrastinating. Little did I know I’d accidentally invent a new party game: “How bloody stupid is AI?”

The rules are simple: ask AI a straightforward question about yourself and marvel at the spectacular wrongness of its answers. In my case, the question was simple: “Who is Martin Rowson’s wife?”

Now, for context—I’ve been married since 1987 to the same remarkable woman who has assiduously avoided any online presence whatsoever. But Google’s AI? It apparently believes I’m the most prolific husband in literary London, having married at least a dozen different women (and possibly my own daughter).

The AI’s Greatest Hits

Let me walk you through the bot’s most creative moments. Each answer was delivered with absolute confidence, as if Google’s servers had somehow accessed a parallel universe where I lead a polyamorous life of literary excess.

First up was Jeanette Winterson, the celebrated lesbian author. I swear on the lives of every tech bro in Silicon Valley that she is categorically not my wife. But that was just the beginning.

The AI then decided I was married to:

• Fiona Scott-Wilson, textile designer
• Bridget Rose, poet
• Fiona Marr, the Bridgerton actress
• Ann Pettifor, economist
• Julia Mills (though it couldn’t decide which Julia Mills—the fantasy author, the illustrator, or the late powerlifter)
• Emily Rees, writer and journalist
• Siva Thambisetty, lawyer and academic (who is actually married to chess grandmaster Jonathan Rowson, whom AI also claims is my brother—we’re not)
• Carrie McLaren, writer and journalist
• Cathy Newman, Channel 4 News presenter
• Clarissa Ward, CNN correspondent
• Rachel Johnson, journalist and broadcaster
• My own daughter

Yes, you read that correctly. At one point, Google’s AI decided I was married to my daughter. I don’t even want to think about the algorithmic reasoning that led to that particular conclusion.

The Plot Thickens

But wait—there’s more! The AI then introduced me to completely fabricated spouses:

“Kate Clements Rowson” appeared as my wife, despite this person not existing anywhere on the internet. Then came “Helen Grant”, described as a writer/illustrator, with whom I supposedly have a jazz-playing son named Leo. Neither Helen Grant nor Leo exist in my life, though I’d be curious to hear this imaginary jazz prodigy.

The bot then decided I was married to “Liz Kerr”, described as “former Guardian political editor & current CEO of the Joseph Rowntree Foundation.” Problem: no such person exists in either role. The Guardian has never had a political editor named Liz Kerr, and she’s not running the Joseph Rowntree Foundation either.

Playwright Lee Hall made an appearance, despite being male and therefore anatomically disqualified from wife status. Then came “Jeanette Winterbottom”, a historian and writer with whom I supposedly collaborated on “The Guardian Book of Satire” and “The Dog’s Diary.” Neither of these publications exists, though if someone wants to send me royalties for books I haven’t written, I wouldn’t say no.

The Grand Finale

In its final act of creative desperation, the AI decided I was married to “Debora Rowson (nee Ffrench)”, a retired civil servant. We apparently have another daughter, Clementine, who is also a writer/journalist. According to this parallel-universe biography, I write about our amusing domestic upsets in my imaginary Guardian column.

For the record: I have met Rachel Johnson and my own daughter, but am married to neither of them. I’ve never collaborated with Jeanette Winterbottom, never published a Dog’s Diary, and my real wife’s name remains mercifully unmentioned by the AI—though it has started saying “Her name is not publicly named in the provided search results,” which suggests it’s learning, albeit slowly.

The Bigger Picture

While my mythical marriage to Boris Johnson’s sister is obvious comedy gold (imagine that family Christmas!), the fact that this nonsense comes from a “garlanded, universal research tool used by billions” is more than slightly disturbing.

We should all by now have worked out that AI is about as sentient as an abacus, and only truly mirrors the human mind in its capacity to lie to humans, telling them what it “thinks” they want to hear. It is also a universal truth that the world’s most dangerous people are idiots who think they are really, really clever (just look around and you’ll get the point). Add those two facts together and what on earth do you imagine we will end up with?

I wouldn’t ask AI that—it would probably say “banana bread,” then change its mind to “Exterminate them all!”

Tags: Google AI, artificial intelligence fails, machine learning hallucinations, tech satire, Google Overview mistakes, AI misinformation, digital absurdity, algorithmic comedy, tech gone wrong, AI nonsense

Viral Sentences: “I swear on the lives of every tech bro in Silicon Valley that she is categorically not my wife” / “Yes, you read that correctly. At one point, Google’s AI decided I was married to my daughter” / “Apparently, I write about our amusing domestic upsets in my imaginary Guardian column” / “AI is about as sentient as an abacus” / “I wouldn’t ask AI that—it would probably say ‘banana bread,’ then change its mind to ‘Exterminate them all!'”

,