Somewhere in a brand team’s inbox this week sits an email from Open.AI that reads like a utility notice. Nobody forwarded it to the creative director. Nobody put it on the agenda. It will be archived by Friday, and on October 30 it will quietly change who writes that brand’s ads.
Here is what it says, according to Barry Schwartz at Search Engine Roundtable, who published it on October 20. Starting Friday, October 30, Open.AI will switch on “creative enhancements” for new and existing self-serve ChatGPT Ads campaigns. By default. The toggle is already on in the settings panel. You don’t opt in. You opt out, or you don’t, and either way you’ve made a decision.
The enhancements are not cosmetic. Open.AI will take your existing ad and your landing page and rewrite the message: new or updated headlines and descriptions written to connect your offer to whatever the person happens to be talking about, translated copy, resized images, and destination URLs swapped to a “more relevant” page on your site. Some of those variations get prepared in advance. Some get generated in real time, inside the conversation. And per the email, the enhanced ads “can run automatically without requiring you to review or approve each version.”
Read that line twice. Then go find the legal team that took eleven days to approve your last Instagram caption.
The same day, October 30, Meta’s revised commercial Terms take effect. In their first draft, businesses that kept using Meta’s platforms would have authorized the company to send scrapers, bots and crawlers to their websites and to periodically pull updates for training and research. Stephen Beale of The Steampunk Explorer flagged the language, legal experts quoted by Nieman Lab said it could justify widespread scraping, and Meta revised the wording over the weekend. A spokesperson told Social Media Today the crawling terms “only applied to our ads and business products.” Which is the point. Your site still feeds the ad machine. The machine just promised to stay in its lane.
So: one platform will write your words on the fly, and another will read your website to help decide who sees them. Same date. Two calendar invites nobody sent.
We are not here to tell you AI is evil. We use it. Every serious shop does. We’re here to say something more uncomfortable: the brand voice you spent years defining is about to become a default setting in someone else’s dashboard, and most brands will find out about it from a screenshot.
Think about what a brand voice actually is. It’s not a PDF with a color palette and three adjectives. It’s a thousand small refusals. The joke you didn’t make. The word you banned because it sounded like a bank. The promise you wouldn’t put in writing because the product couldn’t keep it. Voice is mostly the stuff that never ships. A system optimized to connect your offer to the conversation has no idea what you refused. It only knows what converts.
And here’s the shit the “less work for you” framing skips. The early reviews aren’t glowing. Schwartz’s piece quotes Melissa Mackey saying her team had already paused its ChatGPT Ads test over poor lead quality. Anthony Higman went after the framing directly, arguing it suggests ads may start showing up on queries advertisers never wanted. Neither of them is a Luddite. They’re paid-search people. They know what an automated default does to a budget, because they’ve watched broad match do it for fifteen years.
Now add the research. In September, Harvard Business Review published work from Adam Peruta of Syracuse’s Newhouse School comparing AI-generated and human-made video ads from major brands. Viewers mostly couldn’t tell them apart. The AI ads still did worse on short-term sales potential and on long-term brand equity. Indistinguishable and inferior at the same time, which is the most dangerous combination in marketing. It’s the ad that sails through internal review and then quietly loses money in market.
Consumers, for their part, are less forgiving than the platforms assume. A Talker Research survey commissioned by Auros, released October 19, found that 44% of Americans go out of their way to buy from brands that don’t use AI in their ads, and 53% don’t trust AI-advertised products to be described accurately. That’s not a niche of purists. That’s nearly half the room.
Put those three things on one table. A default that rewrites your copy without per-version approval. Research saying the machine-made version underperforms even when nobody can spot it. And a public that’s actively punishing brands it suspects of letting the robots talk. That’s not an efficiency story. That’s a brand-risk story wearing an ad-ops costume.
Here’s where the smart people in the room push back, and they’re not wrong to.
Personalization is the whole promise of conversational ads, they’ll say. A static headline in a chat interface is a billboard in a living room. If someone is asking ChatGPT how to plan a bachelorette in Palm Springs, the ad that mentions Palm Springs will beat the ad that doesn’t. Real-time rewriting is just relevance at the speed of the conversation. And besides, you can opt out whenever you want. OpenAI says so in the same email: opt out before October 30, or change your settings any time in Ads Manager, with campaign-level settings overriding account defaults.
All true. And none of it touches the real problem.
The real problem is that “you can opt out” is doing an enormous amount of work in that sentence. Defaults win. They win in retirement plans, organ donation forms and cookie banners, and they’ll win in ad accounts. The brands that keep their voice on October 31 won’t be the ones with the best guidelines. They’ll be the ones where a specific human being owned the toggle, read the email, and made a call. Most organizations don’t have that person, because until this month the job didn’t exist.
We’d argue the job is creative direction. Not media buying. Not ad ops. The decision about whether a machine may rewrite your brand in public, in real time, without review, is a creative decision with media consequences, not the other way around. If it gets made by whoever happens to be logged into Ads Manager that afternoon, you’ve outsourced your brand voice to the person with the password.
This is where we think a lot of brands have it backwards. They treat AI tools as a production question: faster cuts, more variants, cheaper resizes. Fine. That’s what the tools are good at. But the moment a platform starts writing new claims, choosing new landing pages and talking to a customer on your behalf, it has crossed from production into authorship. Authorship is the one thing a brand cannot delegate and remain a brand.
So what do you actually do between now and October 30? Not much, which is the beauty of it. Find the person who owns the ChatGPT Ads account, if there is one. Decide on purpose whether enhancements are on, at the account level and per campaign. If you leave them on, write down what the machine is not allowed to say, and go look at what it’s saying every week like your reputation depends on it. If you turn them off, write down why, so the next person doesn’t flip it back to chase a cheaper click. Then check your robots.txt before Meta’s new Terms land, because Meta itself pointed to it as the way you signal your preferences.
None of that is hard. It’s just unglamorous, which is why it won’t happen at most companies until a customer screenshots an ad the brand never wrote.
We’ve spent years telling clients that culture is made by people who refuse things. The DJ who won’t play the request. The chef who won’t serve it with ketchup. The designer who kills the logo everyone in the room liked. Every brand worth following has someone like that in the building, guarding the line between what the brand would say and what it merely could say.
On October 30, the platforms are offering to retire that person.
Don’t let them do it by default.
