OpenAI Just Opened ChatGPT Ads to Every US Business. Here's What Changes for Solo Operators Who Sell Anything.
On May 5, OpenAI removed the $50,000 minimum spend from its Ads Manager and opened self-serve campaign creation to all US businesses. Cost-per-click bidding joined the existing $60 CPM model. A Conversions API and measurement pixel went live. If you have a product that solves a problem someone in the US might type a prompt about, you can now run a campaign inside ChatGPT for the cost of a few hundred dollars of test budget.
The audience: 900 million weekly users as of OpenAI's February 2026 disclosure, up from 800 million in October 2025. The conversion rate advantage: Criteo, the first ad tech partner in ChatGPT's pilot, reported AI-referred users converting at 1.5x other referral channels in March; updated May 2026 data puts select retail categories at approaching 2x traditional search. The moment: right now, before CPCs normalize, before competitors figure out what works, before the playbooks get written and published.
This is the window. It won't last long.
What actually changed on May 5
The original ChatGPT ads pilot ran in closed beta with a $50,000 minimum, effectively closed to anyone who wasn't a large brand with a media agency. The May 5 launch removed that floor entirely. Self-serve creation is now live at ads.openai.com for US businesses.
The changes in one list: no minimum spend, CPC bidding alongside $60 CPM, a Conversions API for measuring post-click outcomes, a measurement pixel for on-site attribution, and access through agency partners (Dentsu, Omnicom, Publicis, WPP) or direct self-serve. Ad tech integrations include Adobe, Criteo, Kargo, Pacvue, and StackAdapt.
International expansion to the UK, Mexico, Brazil, Japan, and South Korea was announced for "coming weeks" after May 7.
Content restrictions exclude dating services, health and medical products, financial services, and political campaigns. More on why that matters for picking your category below.
Context-hints are the skill, not audience targeting
Here's the thing that will separate the operators who win early on ChatGPT Ads from the ones who waste their test budget: this platform doesn't work like Facebook or Google.
On Facebook, you build an audience: demographics, interests, lookalike models, behavioral signals. The targeting layer is the product. On Google, you buy keywords and bid on search intent. On ChatGPT, you write context-hints: natural language descriptions of when your ad should appear. You're describing the conversation, not the person.
A context-hint for a solo developer tool might read: "when a user is asking for help automating repetitive workflow tasks in Python or JavaScript." Or: "when someone is researching alternatives to expensive SaaS project management tools for small teams."
The operators who will perform best are the ones who can articulate, in clear prose, what their buyer sounds like when they're in the moment of need. Not who they are, but what they're saying when the problem is live in front of them. If you've been doing content marketing, SEO, or copywriting for any length of time, you've been building this muscle without knowing it. Your customer research, your FAQ answers, your sales call notes: all of that is training data for writing context-hints that actually convert.
What the 900M user number means for a solo operator
You do not need to reach 900 million people. You need to reach the 0.01% who are your actual buyers at the moment they're typing the prompt that your product answers.
ChatGPT's scale matters for two reasons. First, the absolute number of people typing prompts adjacent to your category is large enough that even micro-targeting has volume. If you sell a tool for startup founders doing their first SOC 2 audit, and even a fraction of a percent of ChatGPT's user base has ever typed "how do I start a SOC 2 audit for my startup," that's tens of thousands of high-intent conversations happening monthly.
Second, scale means the platform will have enough signal to train its matching model quickly. Early advertisers benefit from the system still learning, and from ad fatigue not existing yet. ChatGPT users are not burned out on ads in this interface because the ads have barely existed.
The excluded categories and what they tell you about what works
Dating services, health and medical, financial services, and political campaigns are off the table. These aren't arbitrary restrictions. They reflect where the FTC and EU regulatory frameworks would most likely scrutinize AI-delivered promotional content. OpenAI is being cautious about regulatory exposure.
What this means for category selection: the categories not on the exclusion list have a cleaner runway. Developer tools, SaaS productivity, professional services (consulting, coaching, agency), e-commerce, B2B software, information products, and education are all eligible. If your product lives in one of those spaces, the competitive set on this platform is smaller than you'd expect, because many potential advertisers are still sitting on the sidelines waiting to see whether this works.
The first-mover window
Every new ad platform has a period where early advertisers get disproportionate results before CPCs normalize and competition drives yields down. Google AdWords CPCs in the early 2000s were fractions of what they are today. Facebook Ads in 2012–2014 were generating $0.01 clicks in certain categories. TikTok Ads direct response in 2020 was producing ROAS that made Facebook look expensive.
ChatGPT Ads are at that moment right now.
I don't know how long the window lasts. It could be three months, it could be twelve. It depends on how fast large advertisers ramp, how quickly CPCs respond to competition, and how soon ad-fatigue sets in with users. But the window is open. The platform is self-serve. The minimum is zero.
The honest counter-case: ChatGPT Ads could underdeliver. The context-hints model is new and OpenAI's ad targeting hasn't been battle-tested at the performance marketing level that Meta and Google are. Early CPC data could be misleading if conversion rates aren't holding up downstream. There's a real possibility that the first wave of advertisers writes off the platform after a few thousand dollars of spend that doesn't yield clear ROI.
That's a genuine risk. But it's also the risk that keeps competitors out of the channel while you're learning it.
What I'd actually do
I'd allocate $500–$1,000 as a test budget. I'd write five distinct context-hints, each targeting a different way my buyer might phrase the problem my product solves. I'd run them for two weeks and look at click-through and conversion data before making any judgment.
I would not expect the first campaign to be profitable. Learning how context-hints work in practice is worth $1,000. Figuring out which prompt patterns match my buyers before CPCs normalize is worth $1,000. Being able to say "I've been running ChatGPT Ads since May 2026" in a year when everyone else is just starting is worth $1,000.
The operators who will win this channel are the ones who start now and iterate, not the ones who wait for the case studies.
Author
Lukas
@lukcombinator