AI-generated ads fooled marketing experts and outperformed typical US print ads on a test that measured creativity and potential to spur emotional responses.
AI-generated advertisements proved generally indistinguishable from human-made ads in a competition described as the first ad Turing test.
AI duped advertising experts BrXnd.ai |
The competition was inspired by the original Turing test, which challenges how well machines can mimic humans in conversation. In this case, a panel of 17 marketing experts achieved just 57 per cent accuracy in correctly identifying three ads primarily made using generative AIs and seven ads made by marketing students. The results were announced at the inaugural BrXnd Conference on marketing and AI in New York City on 16 May.
“We put together a jury with probably about 300 years of combined advertising experience – people who have seen many thousands of ads,” says Noah Brier at AI marketing firm BrXnd, who co-organised the ad Turing test.
The competition required non-human-made entries to use AI tools – such as ChatGPT and Stable Diffusion – for creating and assembling all the image and text components. Both the marketing students and the AI teams were asked to create print ads based on the description and logo of a fictional energy drink brand called Volt.
The marking research company System1 Group scored the ads on their creative quality and ability to elicit emotional responses. In the scoring framework, 5.9 stars is the highest ranking and 1 is the lowest, with the typical US print ad earning an average score of 1.8 stars.
One of the AI ads titled “Hustle Hydrated” scored lowest of all the entries with 1.4 stars. But the other two AI ads – “Shock Your System” and “Carpe Diem” – scored 2 and 2.1 stars. Any ad scoring above a 2 is considered to be “outperforming the norm and pretty damn good”, says John Kearon at System1 Group.
At the conference, two experienced marketing professionals used AIs to churn out a new ad every 12 seconds, creating 50 ads in about 10 minutes. They also created an even higher-scoring ad that earned 2.8 stars – but to do that, they broke the rules by using their expertise to lay out the AI-generated images and text.
Companies such as Google and Meta have recently announced plans to offer generative AI tools that can produce images and text for advertisers. Mark D’Arcy at The Brandtech Group, a marketing technology firm, says such ad-generating AI tools could enable more creativity in advertising.
But the risk of bias or copyrighted material creeping into AI-generated ads means humans still need to review the results, according to several conference speakers.
1 Comments
This is an indictment on the intelligence of marketing executives, not an accomplishment for AI models.
ReplyDelete