ChatGPT to Fill Image-Generation Wait Time With Ads, Finally Giving the Loading Bar a Job
OpenAI will test labeled visual ads during ChatGPT image generation later this month in the US, as the chatbot reaches 1.2 billion weekly users and an ad stack with at least ten measurement partners.
Generating an image in ChatGPT involves a short wait while a progress indicator does its best impression of progress. OpenAI has decided that this interval, historically used by humans to reflect on their prompts, is under-monetized. Later this month, the company will begin testing visual ads during image generation in the United States, with an initial group of advertisers.
The ad, per OpenAI as quoted by AndroidHeadlines and iGeeksBlog, appears below the progress indicator while the image renders. It is, in the company’s words, “clearly labeled” and stays “outside the image being generated.” So you can relax: nobody is putting a soda can in your dragon.
The fine print, which is the funniest part when it’s accurate. AI Weekly reports that the ads can show product inspiration, product use or what a service makes possible, and that OpenAI says ads do not influence ChatGPT’s answers. Paid tiers (Plus, Pro, Enterprise) stay ad-free, per both AI Weekly and AndroidHeadlines; Winbuzzer adds that Free and Go users may see ads, and that free users can opt for an “Ads-Free” experience with reduced usage limits. That last sentence deserves a moment. The deal on offer is: pay with attention, or pay with fewer images. Every ad-supported product has this deal. This one just states it with a straight face.
The audience, per OpenAI’s numbers, is 1.2 billion weekly users. AndroidHeadlines pegs ChatGPT at about 2.5 billion prompts a day, against Google’s reported 15 billion daily searches, which it frames as a constraint on ad inventory. In plain terms: OpenAI has a lot of people and not nearly as many opportunities per person, so it is creating opportunities. A loading bar is, structurally, a captive audience with a short fuse.
The plumbing
The most revealing detail is the infrastructure behind the ads. AndroidHeadlines names measurement partners including Kochava, AppsFlyer and Triple Whale; Winbuzzer adds Hightouch, Tealium and LiveRamp, which let advertisers connect purchase and signup data to campaign reports. AI Weekly counts 11 or more such partners. The company has also brought in DoubleVerify and Integral Ad Science for brand-safety work, aimed at keeping ads out of what AI Weekly quotes as “emotionally vulnerable, sensitive, or otherwise unsuitable conversations.” AndroidHeadlines says those partners cannot access private conversations while doing it.
So the product that was introduced as a research lab’s gift to humanity now has attribution pixels. This is not a criticism; it’s a Tuesday. It’s merely striking to see a conversational superintelligence candidate onboard an ad-tech ecosystem with the pace of a mid-sized grocery chain.
How’s it working?
Winbuzzer cites a case study OpenAI shared in which WeightWatchers achieved a cost per acquisition 15.3 percent below its blended paid-search benchmark. It also reports that agencies have seen inconsistencies, including one that saw roughly 100 clicks in ChatGPT while its own tracking recorded about 20. Advertisers also face unpredictable inventory, and some brands have held back conversion tracking because their legal teams were uneasy about the data-sharing arrangements, per the same report.
Five-to-one discrepancies between clicks and tracked clicks are the sort of figure that, in any other medium, prompts a polite email and then a less polite one. Here it’s described as growing pains, which is a fair label for a program still in its test phase.
Where does that leave users? AndroidHeadlines notes that new European rules will require OpenAI to maintain a public ad library starting in January, so at least one jurisdiction will get a catalog of exactly what played while its citizens waited for a cat in a spacesuit.
There’s a broader pattern, too. Every consumer technology that reaches a billion people eventually discovers that the cheapest place to put a message is wherever the user is already waiting. Elevators got screens. Gas pumps got screens. Now the sliver of time between “make me a poster of a corgi CEO” and the corgi CEO has become a media property.
The company says the ads won’t change what ChatGPT tells you, and nothing in the reporting contradicts that. We’ll simply note that the loading bar has been the most honest part of the whole experience, since it was the one thing on screen that never claimed to be anything but a delay. It’s now ad inventory. It was nice while it lasted.
Sources: