OpenAI Just Monetized the Loading Screen
OpenAI just put a price tag on the most boring five seconds in AI: the wait while your image renders. On Monday, October 5, the company announced it will begin testing a visual ad format that appears during ChatGPT image generation, starting later this month in the United States with a limited group of advertisers (Harava News). Dead time was the last moment on earth no platform could sell. OpenAI just proved the inventory was always there. It was just unbilled.
Think about what this actually is. The first wave of ChatGPT ads was polite sponsored units at the bottom of answers: a headline, a description, a link, priced per click. This is something else entirely. Full images, shown while you are already in a visual frame of mind, staring at a screen that is literally assembling something you imagined. Nobody has ever owned that moment, because nobody built a reason to look at it. OpenAI did.
What it is
The new format puts clearly labeled image ads inside the image generation flow. OpenAI says they will appear alongside the generation process, remain separate from the image being created, and never alter what the model produces. Ads will not influence ChatGPT’s answers. Advertisers cannot steer, rank, or touch responses. Conversations stay private from advertisers, who never see them (Thurrott).
The pitch to brands, in OpenAI’s own words: “Our new visual ad format helps people imagine how products and services could fit into their lives. Through images showing product inspiration, product usage, or the experiences they make possible, people can discover a new brand, understand what makes an offering distinctive, and decide what to buy.”
Read that line again. Product inspiration. Product usage. Experiences. OpenAI is not selling clicks here. It is selling the dream right before the render. The creative brief writes itself, and it is nothing like a search ad.
The test starts in the US later in October with a limited advertiser group, before any possible expansion. Ads appear only for users on the Free and Go tiers (Go is $8 a month). Plus, Pro, Business, and Enterprise users never see them. The audience in scope: more than 1.2 billion weekly users, per OpenAI (Thurrott).
What changed
Three things changed, and together they explain why this is happening now.
First, the format ladder. ChatGPT ads launched earlier this year as a single sponsored unit at the bottom of answers, on a cost-per-click basis. In August, OpenAI added product carousels that pull items directly from retailer feeds, a page straight from Google Shopping’s playbook (TechJuice). Now the format moves from the answer to the experience itself. Every step has been a move up the attention hierarchy: from text to images to the creative act itself.
Second, the money has to move. OpenAI announced on September 1 that ChatGPT advertising reached a $1 billion annualized run rate, about 200 days after the first US tests, with tens of thousands of advertisers and ads live in more than 40 countries (Seoul Economic Daily, via CNBC). That is a monster ramp. It is also, by OpenAI’s own internal math, not enough: the company is chasing roughly $2.4 to $2.5 billion in ad revenue for 2026, and the current pace falls short, with one Emarketer analyst calling the milestone “incredibly impressive and terribly disappointing” (CoinCentral). You do not close a gap like that with more text units at the bottom of answers. You close it with premium visual inventory that only you can sell.
Third, the referee role got awkward. Last month OpenAI quietly told ad partners it would no longer accept ads for standalone image and voice generation products, a move first reported by The Information. Adobe, an early pilot partner that had promoted its Firefly image generator inside ChatGPT, found its ads rejected (WebProNews). Video generation ads are still allowed, but the line is drawn exactly where OpenAI’s own products live: image and voice. OpenAI is now the only image-generation company allowed to buy the image-generation moment. Let that sink in.
What works
This is, objectively, one of the most interesting ad placements invented in a decade. Here is why.
The moment is pure intent. Nobody generates an image idly. You are decorating a nursery, planning a wedding, designing a brand, picturing a trip. You are mid-dream, staring at the screen, already thinking in pictures. An advertiser who shows up there with the right imagery is not interrupting a task. They are joining one.
The creative bar is different, and that is the unlock. Search ads reward the best answer. Social ads reward the best hook. This format rewards the best aspiration. Brands that sell outcomes you can picture, travel, fashion, home, food, automotive, wellness, have a native advantage here. A direct-response coupon will look cheap in this slot. A beautiful image of the life your product makes possible will look like it belongs there.
The early numbers suggest the audience buys. When OpenAI announced the $1 billion run rate, it disclosed that one ecommerce advertiser saw a 3x return on ad spend across 28-day campaigns, and that a technology partner saw more than 80 percent of ad-driven ChatGPT traffic come from new customers (AI Stock Wire). These are company claims, not audits, but the direction of travel is consistent: this is a new-customer channel, not a retargeting one.
What breaks or stays fenced
Start with the trust architecture, because OpenAI is clearly nervous about it. The company says ads will be kept out of emotionally vulnerable, sensitive, or otherwise unsuitable contexts (Thurrott). DoubleVerify and Integral Ad Science are running brand-suitability pilots in controlled environments, without access to real conversations (TechJuice). That is the right privacy call. It is also a maturity lag: suitability scoring built without seeing the actual context will improve slower than the open-web equivalent, and the advertiser’s real question, “what was on the screen next to my brand,” has a fuzzier answer here.
Then the competitive question, which is spicier than it looks. OpenAI now blocks rival image and voice AI products from advertising in the very product where it is launching visual ads. Whatever you think of the business logic, the optics are monopoly-adjacent: the landlord is also the biggest advertiser and the only one allowed to sell apartments. Regulators are already sniffing around AI ad practices; this will not make them less curious.
Finally, the creative adjacency risk nobody wants to name. Your brand’s beautiful inspiration image appears while a stranger generates something you cannot predict, in a flow you cannot control, next to a prompt you will never see. OpenAI promises separation and labeling. Separation is not context. The brand-safety story here is still version 0.9, and premium pricing on a v0.9 safety product is a negotiation, not a fact.
Who it is for
Visual direct-to-consumer and ecommerce brands that sell something you can picture: furniture, fashion, beauty, travel, home improvement, automotive, food and beverage. If your creative already works on Pinterest and Instagram, you already have the assets.
Performance teams hunting new-customer channels. The consistent signal from OpenAI’s advertiser disclosures is net-new buyers, not cheap retargeting. If your growth problem is saturation on Meta and Google, this is the test that matters.
Agencies building conversational-ad practices. Paid inside AI answers is now a real line on the media plan. The teams that learn its creative grammar first will own the playbook everyone else licenses.
It is not for regulated categories yet. Pharma, financial services, anything where suitability precision is a legal requirement rather than a preference: wait for the verification pilots to harden. And it is not for brands whose legal team needs to pre-approve every adjacency. This slot cannot promise you that.
What to do this week
- Storyboard for aspiration, not conversion. Pull your three most beautiful lifestyle assets and ask: does this make someone feel something before they think something? If your creative is all offers and CTAs, you are bringing a coupon to a gallery opening.
- Model the unit economics of premium conversational inventory. Take your current blended customer acquisition cost, add a premium, and decide what thin-auction pricing would have to look like to beat it. Set the number before the rep calls you, not after.
- Get in the queue. The test is limited and US-only for now. Ask your agency or OpenAI contact about access, and watch for self-serve expansion the way you would watch any new Google surface.
- Write your adjacency red lines now. Decide in advance what you need from DoubleVerify or Integral Ad Science suitability reporting before you spend a dollar. “We will figure out brand safety later” is how case studies become cautionary tales.
- Pair paid with presence. Ads rent the moment; visibility owns it. If you are not also running the free playbook (what AI assistants say about your category when nobody pays), you are buying a slot in a conversation your brand is absent from.
OpenAI spent years insisting advertising was a last resort, then built a $1 billion ad business in 200 days, and now it is selling the seconds between your imagination and the image. The loading screen was the last unsold space in software. It is sold now. The only question is whether your brand shows up as inspiration or as noise.
