Have you ever driven past an obnoxiously large billboard and thought to yourself, I wish it could talk? Well, me neither. But OpenAI is making this dream a reality.
Last Wednesday, the company announced that it would be “reimagining advertising” on ChatGPT. The platform had already been started testing ads in the United States since the beginning of this year, with the hope “that ads can support broad access to AI and keep the trust that makes ChatGPT valuable.” Those ads were similar to the sponsored recommendations that show up at the top of your Google searches: they’re relevant to the topic you’re searching, which arguably makes them more helpful than your average banner or pop-up. Or creepier. Or both.
OpenAI’s ‘reimagined’ ads, however, take things a full step further. Now, each ad is an AI agent itself. With this new ad model – which will be piloted in the US sometime in the next few weeks – “a user can choose to start a clearly labeled conversation with a business-sponsored agent in ChatGPT.”
A blog post announcing the update provides an example. “I need quick healthy dinners for my family this week,” a user types. ChatGPT generates a response, outlining a general strategy and a few possible dinner ideas. Then, at the bottom of the screen, an advertisement for a company called Heirloom Grocery appears: “Need to plan healthy family dinner fast? Chat with us.” The user clicks on the ad (for some reason) and asks the Heirloom Grocery bot to advise on their dinner plans. The bot responds with an onslaught of branded links. Welcome to the future.
My first impulse is, naturally, to assume that no one wants this, that this is the equivalent of a billboard that talks back: dystopian, in the most annoying way possible. After all, we already can’t stand those customer service chatbots that unhelpfully troubleshoot what’s wrong with your wifi modem (well, I can’t, at least). And we’re exhausted by ads creeping into every aspect of our lives (most recently in Toronto: the sky).
At the same time, if I squint, I can kind of imagine myself clicking on one of these ads. If these sponsored AI agents really are so relevant to whatever you’re searching for, and so user-friendly, I can’t help but wonder whether they’ll feel like ads at all. Maybe the whole process will feel more like matchmaking – helpfully pairing companies with people who want to buy whatever they’re selling.
For an exhausted parent looking to whip up easy weeknight dinners, having ChatGPT generate meal suggestions makes life a bit easier. If ChatGPT can also point you to a grocery-ordering service, which has its own agent that can speak with you about what specific products you should add to your cart, that could make life a lot easier. The best part? You don’t even have to ask. The platform anticipates your need, and the ad agent appears to fulfill it.
Paying for ad space doesn’t make a company trustworthy, and it doesn’t mean their product or service is any good.
Of course, this only holds if the company being advertised is reliable, offers reasonable prices, and isn’t trying to upsell you on every item; the glaring problem is that none of those things are guaranteed. Paying for ad space doesn’t make a company trustworthy, and it doesn’t mean their product or service is any good. We’ve all learned this from the incredible volume of ads that have infiltrated our social media feeds over the last few years. Personally, I’ve almost ordered shampoo that was advertised to me on social media, only to look up reviews and see reports of customers’ hair falling out and scalps being burned. Glad I checked.
But there’s a reason I came so close to making the purchase. Social media advertising is incredibly personalized, and it often mimics the style and feeling of regular posting. Unlike the ads living on banners at the bottom of your screen, social media ads – despite being labelled – often blend seamlessly with the content you’re consuming, especially when you’re swiping or tapping through very quickly. Social media ads look and feel somewhat similar to posts from your friends or your favourite influencers – which, I think, is part of what makes them work.
Unlike Instagram story ads, ChatGPT’s ad agents won’t be fully integrated with other parts of the platform. You will have to click on the ad in order to talk to the agent (at least for now). But if the ad agents communicate with the same helpfully informative tone that ChatGPT uses, I’d expect that, much like with social media advertising, the medium may make the ads feel more trustworthy.
In some ways, this isn’t much different from a conspicuously-placed Coca-Cola bottle in your favourite ‘80s movie. But we don’t look to Hollywood to help us plan our lives, or to tell us what’s happening in the world, or to give us advice. Increasingly, people are turning to ChatGPT for all of these things. The platform currently has over a billion monthly users, and about half of all queries use the chatbot “as an advisor,” meaning users are relying on AI to provide them with accurate information of some kind. Other studies show that consuming news via AI is on the rise, particularly among young people.
All this is concerning enough on its own, given AI’s track record of hallucination. But with advertising agents in the mix, ChatGPT can distract users from the information it’s providing, information which is, at least theoretically, factual. If it were to start redirecting their attention towards the sorts of ads we already see on social media, that would be a major downgrade.
Overall, I’m disappointed, but not surprised. After all, we’ve seen this film before. Nearly every major digital platform either sucks now or seems like it’ll suck eventually, and the oversaturation of advertising is a huge part of that. Tech activist and journalist Cory Doctorow calls this process ‘enshittification,’ and he argues that it unfolds as digital platforms go from prioritizing users to business-customers to, ultimately, shareholders.
In a society increasingly obsessed with convenience, we may be happy to have ChatGPT present us with a grocery-ordering service as soon as we type the word “dinner.”
First, he says, platforms make themselves as appealing as possible, to lock in their customer base. Then, once they have enough users (say, I don’t know, a billion monthly users?), they start to become more ad-friendly, prioritizing the experiences of business-customers above everyday users. In the final stage of the process, platforms become unfriendly to users and businesses alike; they degrade in quality so that costs can be cut and profits increased.
OpenAI is still a fairly young company, and as of yet, it has had too many competitors and growing pains (including, apparently, the threat of human extinction) to have wandered too far down the enshittification path. Still, these new advertising agents represent a definitive step in that direction.
It’s not inconceivable to me that people will use this feature. In a society increasingly obsessed with convenience, we may be happy to have ChatGPT present us with a grocery-ordering service as soon as we type the word 'dinner.' But I am convinced that this feature will make the platform a far worse destination for people trying to learn something without being sold a product.
And what’s next? An ad agent that takes over conversations on the fly? The AI-powered talking billboard of my nightmares? ChatGPT is, after all, early in enshittification’s second stage. We may well look back on this era of chatbot advertising as the good old days.
