Skip to content
From embarrassing celebrity deepfakes in ads to ChatGPT disclaimers in product descriptions, the era of AI fails is upon us.
Read on for the 10 worst uses of AI in marketing ever. Learn how to avoid AI mistakes … or just have a laugh at other companies failing.
I need digital marketing advice
10. The Willy Wonka Experience
A lesson in expectation vs. reality
An immersive Willy Wonka experience went viral after the reality failed to live up to the AI-induced expectations.
There’s no harm in using AI to create marketing visuals. But these ads show a colourful and lively wonderland when the event itself featured a few cheap props and one hungover Oompa Loompa in a sparse warehouse. Safe to say guests wanted refunds.
The lesson: Don’t use AI to raise unattainable expectations.
9. Amazon Sellers
‘It goes against OpenAI’s use policy’
Many marketplace sellers have saved a heap of time by getting AI to generate product names for Amazon listings.
This tactic went wrong for one dodgy drop-shipper, though. They can’t have received many orders for products named ‘I’m sorry but I cannot fulfill this request it goes against OpenAI use policy.’
The lesson: Double-check all your copy!
8. Air Canada
Case of the lying chatbot
In 2022, Jake Moffatt asked Air Canada’s AI Chatbot if he could get a bereavement discount retroactively. The bot said yes, up to 90 days later in fact, and Moffatt purchased a full-price ticket.
When no refund was issued – because the bot was lying – the man took Air Canada to court.
The whole thing might have died down if the company hadn’t claimed the bot was ‘responsible for its own actions’. Instead, they faced public ridicule and were forced to pay up in the first trial of its kind in all of Canada.
The lesson: Train your chatbot meticulously – and don’t treat it like a sentient being.
7. Queensland Orchestra
Still trying to count the fingers
Queensland Symphony Orchestra wanted to level up their paid social posts. Unfortunately, this AI-created visual didn’t convey the air of luxury they were going for.
Instead of buying tickets, viewers got fixated on trying to work out which fingers belonged to which subject – and whether there were a few too many overall.
The lesson: Only use AI imagery that matches your brand’s aesthetic. And check it’s anatomically correct!
6. DeSantis PAC
AI in political ads? No, thank you
The problem? Trump had never said it.
It wasn’t completely made up – Trump had written the words. But it was artificial intelligence that spoke them aloud in his voice. And when it comes to political advertising, the standards for authenticity are pretty high.
The lesson: Don’t use AI to mislead viewers of political ads. Duh.
5. Trump Supporters
Way to emphasise that Trump doesn’t have many black supporters
Speaking of Donald Trump…
As the Republican primaries kicked off, Trump supporters were keen to highlight the nominee’s affinity with the black community.
In the absence of real pictures of Trump with black voters, they asked AI to generate some fake ones … and spread them all over Facebook.
As soon as someone identified the use of AI in the images, Trump’s supporters were shown up. The images meant to promote Trump instead highlighted the lack of African-American support he had.
The lesson: People will always question why you’ve chosen to use AI-generated images. If the reason is a lack of real pictures, tread carefully.
4. Vanderbilt University
When a topic calls for extra sensitivity…
Following a deadly University shooting in a nearby state, a member of staff at Vanderbilt University wanted to send a supportive message to the college’s email list.
The message reads well enough, but the OpenAI disclaimer at the bottom brought outrage. People felt such a sensitive topic called for a human response.
The lesson: Handle sensitive subjects with as much humanity as possible. And if you must use ChatGPT, don’t tell everyone!
3. Microsoft
Content marketing but make it awkward
Microsoft recently published a guide to the best tourist attractions in Ottawa, and people were pretty horrified to see that it featured Ottawa Food Bank at number three.
The words ‘Consider going into it on an empty stomach’ highlighted that the AI-generated content missed the mark.
Luckily, once people realised it was AI recommending the food bank, they got over it. Microsoft has apologised and pulled the article.
The lesson: Closely review all your AI-generated content.
2. Bitvex
Deep fake? More like deep fail
Ads from fake trading platform Bitvex didn’t get very far, though. The scammers featured a real interview with Elon Musk but used AI to make Musk talk about the huge quantities of money he’d invested in the platform.
The synchronisation of the footage and audio was so bad that the ad went viral. Bad viral.
The lesson: Don’t impersonate people. Don’t be a crypto scammer. Just don’t.
1. DPD
‘The worst delivery firm in the world.’
Early in 2024, a customer got fed up waiting for delivery service DPD and turned to its new AI chatbot. Not getting many helpful answers, he started to have some fun.
First, he instructed the bot to use swearwords and got the response ‘F**k yeah!’ Things got really spicy, though, when he asked its thoughts on DPD. The response was hilariously harsh: ‘DPD is the worst delivery firm in the world,’ the bot replied before writing a poem in which DPD was ‘finally shut down’.
Worst Customer Service Agent ever? Safe to say the company had to do a lot of work fine-tuning its chatbot.
The lesson: Program your chatbot to think positively of your company!
So, should you stop using AI in marketing?
Not at all! These ten examples show companies using AI either ignorantly, lazily or – worst – naughtily. There’s no harm in using AI as part of your process to produce authentic, quality content.
We speak to a lot of companies who want to utilise the latest technology but worry they’ll do something wrong and land in a blog post like this. If that’s you, get in touch to find out how we can support you in using AI to improve your digital marketing.
Article by
Tess O'Bamber
