When AI starts writing ads just for you
A learn article on how AI personalizes marketing: predictive recommendations, generative ad copy and visuals, automated A/B testing, and send-time and channel optimization. It cites brand examples such as Papa John's, L'Oréal, Delta, Nutella, and Farfetch, and closes with transparency concerns and EU AI labeling rules.
I came across a little tidbit the other day — and it caught my attention.
Someone received an email whose recommendation was, of all things, the exact pair of shoes they'd searched for in the middle of the night. Their first thought: does the marketing department have someone staying up late just to watch me? Of course not. It's AI that's keeping an eye on me.
That got me thinking about a phrase — what does it even mean for marketing to "hit the mark"?
Put simply, it's the shift from "send the same message to a crowd" to "send each person the one message they actually want to hear."
Let's start with how it used to be done.
Back in the day, marketing leaned on coarse tags like age, gender, and city. Twenty-five, male, lives in a big city — push this. Did it work? Somewhat — but that's about it. You scroll past ten ads a day, and maybe not even one or two actually speak to you.
Not anymore. AI now watches what you buy, what time you browse the site, where you pause and stop scrolling — even the tone you type in. What does precise mean? This is precise. Sometimes it knows better than you do what you want in this exact moment.
But the thing that's truly valuable isn't that. It's prediction.
AI tracks your every move and dares to guess what you'll want next, putting it in front of you a beat early. Take a pizza chain like Papa John's: it teamed up with a cloud provider to roll out a full suite of AI recommendations — reading the time and the context, pushing a combo on Friday game night, and a late-night snack on those evenings when you're wrung out from a hard week. You never said a word; the company had what you wanted ready before you even asked.
You might wonder — does that human touch really sell more? The data is blunt about it. Personalized content can push revenue up by forty percent. Forty percent — not four.
AI isn't just in charge of "guessing what you want." It also handles "writing it for you."
Pair it with generative models, and email subject lines, ad copy, imagery — even short video clips of ten-odd seconds — all assemble themselves. L'Oréal teamed up with Nvidia to generate ad visuals with AI, and to recommend products based on each person's skin and hair type. Work that used to take an entire design team weeks now flows down a single pipeline.
Even "testing" is something AI volunteered to take over.
A/B testing used to be done by hand — slow going, one attempt at a time. Now, with reinforcement learning, AI generates a whole batch of versions on the fly, switches between them automatically, and keeps whichever one earns the higher click-through rate. I don't have to guess which headline is better — the machine grades my answers every single day.
Then there's the timing of when a message goes out.
Which hour gets the highest open rate? Every answer is different. Some people check their inbox in the morning, some before bed. AI locks onto "when you're most active" and sends at exactly that moment. Send it right, and the open rate climbs on its own — and people feel less bombarded, less annoyed.
It'll even pick its channel. You prefer email? It sends email. You scroll short video? It finds you there. Can't stand typing? It sends a chatbot over.
Go one layer deeper, and some companies have gone so far as to build a "digital twin" of you.
A digital twin is a virtual model of you, built into a computer, that simulates how you might react — and only then does it make a decision. Delta Air Lines did exactly this for its sponsorship of the 2024 Paris Olympics: it ran neural-network simulations of how different customers might pay, then itemized, one by one, the $30 million-plus in ticket sales the campaign drove. Run the buyer through the computer first — then decide how to spend the money.
When it comes to "one of one," some brands have pushed it to the extreme.
That hazelnut-spread brand, Nutella, had AI draw seven million jar labels — seven million, every single jar one of a kind, released in a limited run that sold out almost instantly. Meanwhile Farfetch, the light-luxury e-commerce player, used AI to tweak its email subject lines and content, and its open and click-through rates ticked up by about 7% almost effortlessly.
You might put all this down to brand stunts. But the giants already treat it as infrastructure. Meta, Google, and Amazon all push tools that auto-generate ads and auto-place ad spend — and a line of top luxury brands is queuing up to get in. Personalization is turning from a "nice extra" into the new baseline.
Which is where I have to stop and pour a little cold water on the whole thing.
When I get understood this thoroughly, a string inside me gets pulled — are you understanding me, or seeing through me? There's a line we can't cross: be transparent, give me choices, don't hoard data you shouldn't, and don't use a machine to cut me. Over in the EU, there's a law in the works that would require AI-generated marketing content to clearly label itself as such. Technology can be smart, but it can't be sneaky.
Some sources predict that by 2029, nearly nine-tenths of all ad revenue will fall under AI's influence. By then, no one's asking "should marketing use AI anymore" — the question is "can you use it — and can you use it responsibly?"
When all is said and done, the essence of this business hasn't changed. It's not about treating people as a button to click. It's about understanding a person, and then earning their trust.
When you can give each person a message that's right — and that was written with care — you're not selling them something; you're talking to a person.
And the ones who bother to speak with care end up being the ones who get remembered.