For most of my career, the media decisions belonged to the marketer. You chose the audience, the format, the timing, the split of the budget. The platform handed you levers and you pulled them. That arrangement is quietly changing, and this week gave me three examples worth sitting with.
Start with OpenAI. It's rolling out a product carousel format for ads inside ChatGPT, letting several products from one retailer appear in a single placement instead of the single-product unit it has run for about five months. Digiday reported the detail that stayed with me. OpenAI's system decides whether to serve one ad or a carousel. That choice used to belong to the advertiser. Now it sits with the platform, which pulls product data straight from retailer feeds and makes the call.
Reddit is doing a version of the same thing at larger scale. It has turned community conversations into an advertising signal, using AI to match ads to the context of a discussion, generate Reddit-specific creative, and automate campaign decisions. Reddit says its Max campaigns cut cost per acquisition by 15 percent and raised result volume by 28 percent, and that a feature called Redditor Highlights lifted clickthrough by roughly 11 percent. Those are Reddit's own figures, so I'd treat them as direction rather than proof. The direction is still clear. The company reached $762 million in advertising revenue and about 70 percent year-over-year growth in active advertisers, and much of the optimization now happens without a person touching it.
The third example is smaller and closer to the ground. A study of more than 22,000 follower interactions found that people who received a labelled AI reply to their comment went on to comment 41 percent more often and repost 43 percent more often than people who received no reply. The effect was strongest when the reply arrived within 45 minutes and actually fit the conversation. Speed and relevance at that volume is not something a person does. A system does it.
I don't read any of this as the end of the marketer's job. I read it as a change in where the job sits. When the platform decides the format and who sees it, the work that's left is deciding what you'll let it decide. Which products belong in that feed, and what your brand should never say in an automated reply at two in the morning. Those are judgment calls, and they don't get easier as the automation improves. They get more consequential, because one setting now moves thousands of impressions.
The harder part is that the levers are disappearing while this happens. You can't hand-pick the ad format if the platform keeps that choice for itself. So the skill becomes knowing the system well enough to set it up right, then reading the outputs honestly enough to notice when it's optimizing toward something you never asked for.
This fall I'm sitting down with marketing practitioners here in Vancouver to hear how they're handling this in the accounts they run day to day, not in theory. I suspect the honest answers are messier than the case studies. I'm looking forward to them.
Sources
- OpenAI brings product carousels to ChatGPT ads (Digiday): digiday.com
- How Reddit uses community conversations to improve ad targeting (MarketingTech News): marketingtechnews.net
- Use AI bots to reply to followers (Science Says): sciencesays.com