On February 9, 2026, OpenAI turned on ads inside ChatGPT for Free and Go tier users in the US. Eighty-six days later, on May 5, it opened self-serve access to any advertiser through ads.openai.com, removed the previous $50,000 minimum spend requirement, and added CPC bidding alongside the original CPM model. A platform that started as a closed enterprise pilot became accessible to any marketer with a $25 daily budget in less than three months.
That speed of opening is itself a signal worth paying attention to.
What the platform is and isn't
ChatGPT ads appear after the AI-generated answer, not inside it. They're labeled as Sponsored, separated from the model's response, and served based on conversational context rather than keywords. The user never searched for anything. They were in the middle of a conversation, and the ad showed up because the topic was relevant.
According to OpenAI's own documentation, ads don't appear near sensitive topics including health, finance, dating, or politics. They don't appear in accounts where users are predicted to be under 18. They don't appear in Temporary Chats or after image generations. And they never appear for Plus, Pro, Team, Business, or Enterprise subscribers. That last point is structurally important: the audience is Free and Go tier users only, which is the majority of ChatGPT's 900 million weekly active users, but not the highest-intent paying segment.
As of July 2026, OpenAI also updated the ad card format to a more compact design with larger visual elements, easier to read on both desktop and mobile.
How the buying works
| Launch (Feb 2026) | Self-serve (May 2026) | Current (Aug 2026) | |
|---|---|---|---|
| Access | Closed pilot, invite only | Open to all US advertisers | Open, rolling out globally |
| Minimum spend | $200,000 | Removed | $25/day minimum budget |
| Bidding | CPM only (~$60) | CPM + CPC added | CPM (~$25 observed), CPC ($3-5 recommended) |
| Measurement | No pixel | OAIQ pixel + Conversions API | OAIQ pixel, CAPI, GA4 integration |
| Agency partners | Dentsu, Omnicom, WPP | + Publicis, Adobe, Criteo, StackAdapt | Expanding |
The three campaign objectives available in Ads Manager are Reach, Clicks, and Conversions. Once selected, the objective can't be changed, so it's worth being deliberate before launching. The OAIQ pixel, OpenAI's browser-side tracking tag, sets a first-party cookie on the advertiser's domain with a 30-day attribution window. Because it's first-party and not third-party, it isn't subject to the same browser restrictions that have degraded cross-site tracking on other platforms.
What the performance data says
This is where the picture gets more nuanced. OpenAI has published no cross-advertiser benchmarks for CTR, conversion rate, CPA, or ROAS. What exists is early data from practitioners who were in the beta.
Similarweb's May 2026 analysis puts the overall ChatGPT ads CTR at 0.68%, with the top quartile of advertisers reaching 1% and the best-performing brands hitting 1.57%, with an observed peak of 5.4%. For context, search ads across Google and Microsoft average 6.66%. Display averages 0.1%. ChatGPT sits somewhere in between, closer to display in raw CTR but with meaningfully different downstream behavior.
Real account data from Cogwheel Marketing, published July 30, 2026, shows CTR averaging between 1.6% and 1.9%, materially higher than display, and a 2.2% conversion rate from total impressions over two months of live campaigns in the hotel vertical. Synter's June 2026 account report documented 129,663 impressions and $4,428 in spend across five markets from a single live advertiser. Early data from Criteo shows referrals from LLM platforms convert at 1.5 times the rate of other digital channels, with visitors spending 60 to 80% more time on-site than social traffic.
| Metric | ChatGPT Ads (observed) | Google Search Ads | Display Ads |
|---|---|---|---|
| Average CTR | 0.68% — 1.9% (vertical dependent) | 6.66% | ~0.1% |
| Observed peak CTR | 5.4% | Varies | Varies |
| Post-click time on site | 60-80% higher than social | High | Low |
| Conversion rate vs other channels | 1.5x (Criteo, LLM referral data) | Benchmark | Below benchmark |
| CPM | ~$25 observed | Variable | $2-5 typical |
| Recommended CPC | $3-5 | Variable by vertical | N/A |
Sources: Similarweb May 2026, Cogwheel Jul 2026, Criteo Feb 2026, industry benchmarks.
The pattern is consistent with what AI search data shows more broadly: lower volume, higher intent. The user who clicked was already in the middle of a relevant conversation.
One number that frames the opportunity correctly: OpenAI reported $100 million in advertising revenue in its first six weeks. That's not a test. That's a market validating demand.
The measurement gap
The honest part of this field guide is that attribution is still unresolved in meaningful ways.
The OAIQ pixel and Conversions API give you browser-side and server-side event tracking. But the platform doesn't expose user-level data: who saw the ad, what they asked ChatGPT, or any demographic detail. Advertisers receive aggregated performance data only. That's by design, and it's the right call for user trust, but it limits the optimization levers available compared to Google or Meta.
There's also what practitioners call the conversation gap: a user sees your brand cited in a ChatGPT response, closes the app, and searches for you directly three days later. That conversion shows up in GA4 as branded search or direct. ChatGPT gets no credit. As of April-May 2026, Ads Manager reports impressions, clicks, CTR, CPC, and conversions via the OAIQ pixel or Conversions API, but attribution windows are configurable and not yet fully validated across verticals.
The practical proxy: branded search volume in Google Search Console is your best signal for the broader awareness impact of ChatGPT ads beyond the attributed click.
A channel doesn't work in isolation
There's a pattern in how teams are approaching ChatGPT ads that's worth naming, because it tends to produce disappointing results.
The pattern is treating it as a standalone channel. Launch a campaign, watch the dashboard, evaluate in isolation. That's how most teams evaluated TikTok ads when they launched. And Facebook ads before that. And programmatic before that. Each time, the teams that got the most out of a new channel were the ones that built it into a coherent system, not the ones that tested it in parallel to everything else without a connecting strategy.
ChatGPT ads don't operate in isolation any more than any other channel does. The user who sees your ad in a ChatGPT conversation may have first encountered your brand in a Perplexity citation three weeks earlier. They may have Googled you after the conversation and converted through branded search. The paid placement is one touchpoint in a journey that started somewhere in AI search and ends somewhere in your funnel. If you're running ads without a GEO strategy, you're paying to capture demand that your organic AI presence should be building. If you're running GEO without paid, you're leaving visibility on the table in a channel where your competitors are buying it.
The teams seeing the most traction with ChatGPT ads in 2026 are the ones treating AI visibility as a system. GEO builds organic presence in the moments before the buyer forms their shortlist. ChatGPT ads capture that buyer when they're already in a relevant conversation. Your site converts someone who arrives more informed and more decisive than a typical search visitor. Each layer supports the others. Without the GEO layer, the paid placement has no organic foundation to reinforce. Without the paid layer, you're only visible when the model decides to cite you. Without a site designed for someone who already researched, you're wasting both.
| Layer | What it does | Without it |
|---|---|---|
| GEO (organic AI visibility) | Builds brand presence in AI responses before buyer forms shortlist | Paid has no organic foundation to reinforce |
| ChatGPT Ads (paid) | Captures buyer during a relevant conversation | Only visible when model decides to cite you |
| Site designed for validation mode | Converts someone who arrives already informed | Both layers leak at the bottom of the funnel |
| Branded search monitoring | Measures indirect impact of AI visibility | No signal on what's actually working |
That's not an argument against testing ChatGPT ads. It's an argument for testing them as part of something, not instead of something.
Who should test it now and who should wait
The platform is early. The buying controls are limited compared to what Google and Meta offer. There are no audience targeting options beyond conversational context. No retargeting. No lookalike modeling. You're targeting intent signals in a conversation, not a profile built from years of behavioral data.
That's a different kind of targeting, not a worse one, but it requires a different brief. The creative has to be relevant to a conversation in progress. The offer has to make sense for someone who was already exploring a topic. The landing page has to serve someone who arrives more informed than a typical display visitor.
OpenAI's self-serve launch on May 5 removed the $50,000 barrier, making meaningful tests accessible to mid-market and SMB advertisers for the first time. At $25 daily minimum, a month-long test costs $750. That's a reasonable signal budget for any team that wants to understand the channel before it scales. The window where this is still uncrowded is not unlimited.