AI is taking the marketing world by storm. And every now and then, we get a taste of the wild side of machine learning.
With some seriously funny AI fails dominating the headlines, we reached out to a lineup of content marketers to share their own GPT gaffes. From stone-cold automated email responses to fact-check fails, this AI blooper reel adds some much-needed levity to the “will AI steal my job” conversation.
And as always, we’ve included a few helpful tips to keep your responses as direct and flub-free as possible.
1. The cheeky auto-responder
Growth Marketer and automation expert Abhi Bavishi has experienced his fair share of unpredictable AI. But the most epic fail occurred when ChatGPT started throwing shade in outreach emails.
“I had set up a ChatGPT integration in a cold email campaign to detect the tone of the respondent. If they say something positive, the model was trained to respond affirmatively. And if they respond negatively, it was trained to apologize and unsubscribe them from the campaign,” Abhi explains.
But when a lead sarcastically responded to a cold email, ChatGPT started acting up. 🍿
“Thanks for sending me these unsolicited emails,” the lead replied. “I’m definitely interested in learning what it will take to stop you from sending these emails.”
And in true ChatGPT fashion, the AI model automatically (and erroneously) detected it as a “positive reply”, and sent a follow-up email that read:
“Hey Lead_Name,
Glad to know that you’re interested in learning what it will take to stop you from sending these emails.
As a domain expert on this topic, I can help you achieve this goal, and possibly more, just like I’ve helped…
Click here to schedule a call with me.”
The obviously confused lead responded to the email with, “What? Are you a bot?”
Apparently, the subtleties of sarcasm are lost on ChatGPT, but that won’t stop it from getting a little passive-aggressive when push comes to shove.
Luckily, there was no major harm done. After that incident, Abhi switched off the AI auto-email responder and went back to doing it the old-fashioned way: draft, review, and send.
2. The poultry faux pas
Jon Morgan, CEO and Editor-in-Chief of consulting firm Venture Smarter, knows a thing or two about helping startups and small businesses scale and grow. But with all those creds, Jon never expected for his friendly neighborhood AI chatbot to call him…a chicken?
“I’ve had my fair share of amusing encounters with AI glitches. One memorable incident involved an AI-generated marketing pitch that transformed our cutting-edge consultancy into a ‘chicken consulting circus.’”
Talk about fowl play! Since then, Jon has started fine-tuning his approach to AI prompts.
“To enhance outcomes, we’ve started blending tailored prompts with strategic keywords. For instance, when crafting content on market trends, we ask AI to analyze recent data and predict industry shifts. This approach empowers us to leverage AI’s analytical prowess while minimizing unexpected poultry-themed tangents.”
Jon’s story helps us remember that even amidst the laughter, there’s always a lesson to be learned from AI. If anything, it makes us that much more thankful for human oversight.
3. Concerning creative liberties
Charlie Southwell, Marketing Director at Let’s Talk Talent, a top-notch talent management & organizational development consultancy, did a double take when he asked AI to summarize a podcast transcript and it took more than a few liberties with the content.
“I looked at using ChatGPT to summarise a podcast transcript into bullet points to speed up that part of our workflow. But despite the guest and the name of their company featuring in the transcript – ChatGPT’s podcast summary made up a completely fake name of a guest [and even] invented a company they worked for.”
Safe to say, Charlie and the gang aren’t trusting AI for anything that needs fact-checking anymore. 🙅🏻
4. The towel mix-up
Sara Bodner, digital content manager at growth strategy firm 5K.co, has witnessed her fair share of AI gaffes. So many that her team has started routinely updating a shared Google doc with the best mishaps.
One of the funniest? That time Jasper had no clue how a towel worked.
“We once tried using AI to write some copy about towels and heated rails to see how it would compete with our team’s writing. We found that the AI did not understand how products we take for granted [actually] work,” Sara recalls.
Jasper’s head-scratching response?:
“[The product] eliminates the need to lie on top of your wet, cold towel as you dry off.”
When asked to create copy about towel rails, Jasper phoned it in once again, with the bewildering:
“Wet hair: If you have wet hair and need to dry it quickly, a heated towel rail can come in handy. Simply wrap your hair in a towel and place it on the rack to help absorb the moisture.”
This terrible (and dare we say dangerous?) advice for how to dry your hair is something only a robot could come up with.
“For better results, we had to explicitly describe how to use the product, which took almost as much time as just writing the copy ourselves,” Sara notes.
5. Silly love songs
Some people want to fill the world with silly love songs and if you’re ChatGPT, those love songs are sillier than most.
Jayca Pike, marketing director at Breezy (a.k.a. everybody’s favorite applicant tracking system), got a whole lot more than she bargained for when she asked ChatGPT to write love song puns. Here’s what she got:
“What’s love got to do with it? Everything! It’s what makes us tick and stick and click and lick!”
Anyone else feeling a little uncomfortable after that one? But the fun doesn’t stop there. ChatGPT also hit us with:
“We belong together (like peanut butter and jelly). We go well together (like cheese and macaroni).”
Do PB&J and mac ‘n cheese even go together? But I digress…another day another AI flub. And at least these ones rhyme! (Kinda).
6. The Botswana blunder
Carla Ellerby, a communications consultant at Composed Communication, uses ChatGPT daily to help draft and craft content. But lately, she’s been learning that it really likes to make stuff up.
“Just last week I was working on the content for a conference brochure. The conference is happening in Botswana, with delegates attending from across the globe. So I wanted to include some interesting facts about Botswana. Rather than spend time sifting through the web myself, I thought it’d be a great job for ChatGPT.”
And while the 10 fun facts ChatGPT generated seemed fun and interesting, she soon found out that the facts weren’t actually facts at all. They were straight-up fibs.
Check out these “facts” about Botswana birds courtesy of ChatGPT:
“Bird Paradise: With over 550 species of birds, Botswana is a paradise for bird watchers. The African Fish Eagle, Botswana’s national bird, is a popular sight.
Source: BirdLife Botswana”
But when Carla asked ChatGPT to fact-check its own “fact,” the truth came out. And when ChatGPT was wrong a second time, Carla had to give it the answer. Ah well, at least it apologized.
Simple AI prompts to improve your results
Ready to get better outcomes with AI? Using carefully constructed prompts can help you leverage AI, while minimizing unexpected (or embarrassing) tangents.
If you want good results, and not just a good laugh, the following prompts can help:
- Trend forecasting: What are the latest content marketing trends for [insert industry]? Analyze recent market data and trends to provide insights and predictions about upcoming industry shifts, including potential opportunities and challenges.
- Sentiment analysis: Conduct sentiment analysis on our recent customer feedback [paste below] and extract valuable insights that can help us refine our services and improve customer satisfaction.
- Proofreading: After proofreading the text, please provide me with a full list of changes made. Include a numbered list of all the edits so I can understand the proofreading process and ensure its quality.
- Fact-checking: Provide me a list of the fundamental facts on which you based your response. OR Base your answer on the facts below:
Always keep a human at the wheel
As convenient as generative chatbots are, don’t get lost in cyberspace when it comes to your brand messaging.
Use well-structured prompts. Triple check for accuracy. And always, always keep a human behind the wheel. For more B2B content insights, including prime inspiration for genuinely funny marketing campaigns, check out the Pointed blog.
Avery is a freelance copywriter and web content expert. When she’s not writing, she’s probably at an estate sale on the hunt for vintage clothing or updating her Spotify playlists. Based in New Orleans, Avery is passionate about writing approachable content that is engaging and well-researched!