Ryan Patterson started his own AI-powered production house in Toronto. He is given a script and deploys AI to create images and animations to show to clients.Fred Lum/The Globe and Mail
Advertising is all about parcelling out the right information in the right way to the right people at the right time. As such, there are things advertisers would prefer not to disclose. For some, that includes the use of generative artificial intelligence in finished campaigns.
Some brands in Canada are deploying generative AI, industry experts say, but in subtle ways, and usually on the down-low. “A lot of brands are testing it out, seeing if consumers will notice,” said Jil Lohnes, president of connected services at Salt XC, a marketing agency in Toronto. “I can’t reinforce enough the risk that brands feel about diving in.”
Those dangers are plentiful. Generative AI has proven distasteful to some parts of the public, for one thing, meaning ads made with the technology could attract the wrong kind of attention. Because systems that create images and videos from text prompts are built on content pulled from the internet, users also run the risk of infringing on someone else’s intellectual property. (Some AI companies, such as Moonvalley, say they train only on licensed content and that their models are commercially safe.)
AI can make a mockery of the creative process, too. Generative AI tools tend to default to the most common visual tropes, which is a recipe for blandness and advertising failure. “Brands live and die by their distinctiveness, and many generative models tend to produce homogeneous, undifferentiated outputs,” said Ari Elkouby, chief creative officer at Dentsu Creative Canada, though he added the randomness can produce “happy accidents,” too. While AI programs have improved dramatically over the past couple of years, the output can still be unpredictable and inconsistent, making the process more time-consuming than expected.
Some companies that have gone all-in with an AI-generated ad have faced the wrath of consumers. McDonald’s in the Netherlands pulled one such ad in December after intense criticism, saying the ordeal marked “an important learning” for the company. Coca-Cola’s AI-generated holiday ad last winter, though less hallucinogenic than its attempt in 2024, still garnered online backlash. (Sample YouTube comment: “I’ve never wanted a Pepsi so badly in my entire life.”) Alcool NB Liquor in New Brunswick was similarly pilloried for an amateurish holiday television spot featuring wonky AI-generated people, prompting the provincial minister responsible for the agency to request a halt to paid promotions.
“If the ad is getting attention because of how it was made, that’s a disservice to the brand,” said Patrick Henderson, vice-president of strategy at Mint, an agency in Toronto.
AI superusers are pulling ahead and widening the productivity gap
Given these episodes, it is not surprising that a common occurrence reporting this story went as follows: An agency would reference their use of AI in an ad – typically for minor visual enhancements such as swapping out a background – but then fail to secure permission from the client to discuss it.
An exception is when the technology is integral to the ad itself. Salt worked on a campaign with Visa that involved interviewing girls who play sports about their aspirations, and then used AI to depict them as adults working as professional players, coaches and so on. “It’s not how you would traditionally think of generative AI,” Ms. Lohnes said. “That one’s a bit safer.”
The use of AI during pre-production to create storyboards and develop concepts has proved less controversial. Agencies can present the material to clients, collect feedback and iterate to arrive at a finished idea faster than in the past. “We can visualize things that would have taken weeks to make otherwise,” Henderson said. “We can create 50 versions of it at that moment.”
The trend is creating new opportunities for professionals who know how to use AI tools such as Midjourney, Runway, Adobe’s product suite and many others.
Ryan Patterson, for one, started AI-powered production house Queen One Studios in Toronto in 2024. He and his co-founder Alana Robshaw have used generative AI to create re-enactment footage for a documentary and are taking on advertising work, too. Mr. Patterson will be given a script and then deploy AI to create images and animations to show to clients. “The beauty of using AI is you can get realistic-looking actors in the scenes with consistent characters, and show those emotions,” he said. “It’s easier for the client to understand how a spot is going to land.”
Typically none of this content sees the light of day, however; a real production crew with real actors will make the finished product. “When they film in a studio, they’ll have the reference, and they’ll know the actor needs to look just like that,” he said.
Is our current form of politics capable of governing AI?
When it comes to putting AI-generated content into the world, many brands are starting small, with, say, an image to accompany a social-media post. Often companies are not paying to promote these posts on social networks because that risks exposing it to a wider audience instead of just loyal followers, Ms. Lohnes said. It’s also common for direct-to-consumer product companies to take a single photo of a product and use AI to generate multiple images of it from different angles or settings, rather than having to take multiple pictures.
Some studies show that brands have good reasons not to broadcast their use of generative AI. Research firm Kantar found that 41 per cent of consumers surveyed said that AI-generated ads bothered them, compared with 29 per cent of marketers. It also found that ads containing obviously AI-generated visuals performed worse than those that were subtle. “Gen AI ads are more likely to make people feel, but not always in a good way,” according to the survey results.
Researchers at New York University and Emory University, meanwhile, found that disclosing the use of AI in an advertisement reduced its effectiveness. Click-through rates fell by as much as 31.5 per cent. Somewhat disconcertingly, the study also found that entirely AI-generated ads performed better than those made by human experts, provided consumers weren’t told about the AI part.
Even with the potential pitfalls, some companies are asking agencies to use the technology in hopes of saving money. “There’s this belief that AI means we don’t have to work as hard or as long, and therefore they shouldn’t have to pay as much,” said Ryan Timms, who was president of McCann Canada for seven years until stepping down earlier this year. Some clients came to the agency with a smaller budget but expecting the same quality of work in the same timeframe that a larger sum would command. “You can’t just press a button and have this thing spit out a great product,” Mr. Timms said.
Mo Bofill, partner at creative agency One Twenty Three West in Toronto, has been fielding similar queries. “We’ve had some clients that will question, ‘Do we need a photographer and illustrator to do that, or can AI do that?’” she said. She’s clear with clients that using AI to mimic an artist’s style amounts to taking their intellectual property.
Among Canada’s lawyers, the jury is still out on AI
There is a general push, however, to compress timelines and costs. Occasionally, the production companies the agency works with will say that a special effect can be completed with AI for less money than with traditional CGI methods, but Ms. Bofill said the cost savings have been minimal so far. At least one spot for a brand spearheaded by One Twenty Three West incorporated a generative AI element – but, no, she cannot name it.
Still, there are many industry players who believe it is inevitable that more AI-generated content will make its way into advertisements, especially as the technology improves. “At the end of the day, it will probably be a financial decision,” Mr. Patterson said.
The key is how it will be used, and how advertisers balance consumer perceptions. “If you can tell that it’s AI, then it’s not good enough,” said Stephan Pretorius, chief technology officer at ad giant WPP in London. Consumers don’t want to be duped, he continued, which makes the depiction of AI-generated people in advertisements dicey. That’s also where AI can be a dead giveaway.
But that may not be true for much longer. The technology is getting better, and so are the people using it. Benjamin Benichou founded AI production studio 3.11LABS in New York last year, and in March put out a 30-second ad for U.S. software company RingCentral Inc. Everything, including the voices, are AI-generated. The spot features a harried unicorn, styled and animated to resemble a real-world puppet, working at a flower shop and dealing with a customer. Four people worked on it, whereas a traditional shoot could have required a crew of 20 to 40.
What’s notable is how much more realistic it looks than previous AI ads. “We don’t want anyone to be able to say it looks terrible,” Mr. Benichou said. “That’s the first thing that will make our clients more comfortable going full-on with AI production.”
He doesn’t believe that live-action shoots will cease for advertisements, but AI-based productions will become more common. That will lead to fewer opportunities for people whose jobs are tied to physical productions. “In terms of cost and efficiency, the vast majority of the time, it doesn’t make sense to continue spending millions of dollars,” he said.