If you ask AI for 10 weird campaign ideas, you’ll probably get 10 variations of something you’ve already seen.
So, what makes weird actually work? I spoke with people across marketing, advertising and AI to find out.
Their responses came down to five rules:
Start with the human truth.
Dan Gildoni, founder of Gildoni Ltd., says not to start by asking AI to be weird. Start by giving it something real to work with.
“Give it a contradiction you’ve lived. A category convention you reject. A constraint that looks absurd,” says Gildoni, who works on how AI changes judgment and work.
“Good weirdness is a compressed truth,” he says, “not a costume.”
Abi Hill, founder and CEO of Just Starting Out, a UK platform for new and early-stage business owners, makes a similar point. She uses AI regularly but distinguishes between using it to execute an idea and asking it to invent the idea.
“The weird ideas still win because they interrupt the pattern,” she says. “The best unconventional ideas have something human underneath them, such as “an observation, joke, frustration, contradiction or truth that makes somebody stop and think, ‘Oh God, yes.’”
Weird, but make it yours.
Morissa Schwartz, founder of GenZ Publishing, an independent publisher that focuses on emerging and underrepresented writers, says a “weird” idea works when the strangeness is traceable to the brand.
She offered a simple three-question test:
-
Could a competitor run this unchanged?
-
Does the idea reveal something the brand actually believes?
-
Will people retell it without the ad in front of them?
That first question matters even more when everyone has access to similar AI tools. Polish alone is no longer much of a competitive advantage.
Dawn McGrath, marketing director at Keller-Heartt Oil Co., offered a particularly concrete example from the industrial B2B world.
Her company created “Two Brians,” videos featuring teenager Brian McGrath Jr. asking his father, Brian McGrath Sr., the company's owner and president, basic questions about products such as hydraulic oil.
The format is unusual, but the weirdness has a purpose: Brian Jr. can ask questions a buyer might otherwise feel embarrassed to ask.
The real father-son relationship also makes the format difficult to replicate.
“AI can fake a spokesperson; it can't fake a relationship,” McGrath says.
Ask AI for the obvious, then look for what it missed.
AI is very good at generating plausible ideas.
That’s exactly why marketers need to get better at rejecting them.
Volodymyr Lebedenko, head of marketing and SEO at HostZealot, suggests deliberately generating the obvious concepts first and then asking what assumption they all share.
If every competitor explains a technical product through features, for example, the more interesting opportunity might be to build a campaign around a customer mistake.
“If the audience thinks that whatever you are sharing is strangely accurate, you have something,” Lebedenko says. “However, if they only think your content is strange, you have a gimmick.”
If production gets cheaper, take bigger creative risks.
AI’s value isn’t just generating ideas. It can also make experimentation cheaper.
Nebojsa Savicic, co-founder of Plainly Videos, a company that automates video versioning, sees a trap in that efficiency: when producing variations becomes cheap, marketers can end up making “hundreds of marginally different variants of the same safe idea.”
He calls the distinction “production experimentation ≠ creative experimentation.”
His advice: Use that capacity to make a few changes that are genuinely uncomfortable. Try unusual pacing, an unexpected opening, drastically different framing or an idea the team isn’t sure will work.
The point isn’t to use AI to create endless content. It’s to use cheaper production to make smarter creative bets.
Make the weirdness earn its place.
The final test: Once the surprise wears off, what’s left?
Ed Tsue, Global Chief Strategy Officer at WPP Media, argues that “weird” works when it is both beautiful and confusing.
Tsue says consumers have become so familiar with formulaic marketing — product demos, branding and “reasons to believe” — that they tune it out.
Weirdness disrupts that pattern, creating confusion that drives attention and curiosity. He points to MailChimp’s “Mail Shrimp,” Cadbury’s Gorilla and Thai advertising as examples.
Kasra Dash, founder of The Masterminders, puts another practical constraint on the idea: an unusual concept still needs to be understandable.
His approach is to look outside your industry for formats, trends and behaviors that are already attracting attention, then translate them into a context where your audience has not seen them before.
A strange idea shouldn’t require the audience to spend 30 seconds figuring out the joke or why it has anything to do with the brand.
The best version creates a quick reaction: “I haven’t seen this before.” Followed by: “I get it.”
Open Tabs

This week's email was brought to you by Sonal Nain. Editing by Laura M. Browning.
SUBSCRIBE TO OUR OTHER NEWSLETTERS
Business news, but not boring: The Hustle | Stay current on AI: Mindstream | Grow your business: Starter Story | Sell better: The Science of Scaling
Follow Marketing Against the Grain on LinkedIn and Instagram
Was this email forwarded to you? Sign up here.