Just got into the OpenAI Ads Manager beta after about a 15-day wait. Wanted to share my first experience setting up a campaign because there aren't many hands-on walkthroughs yet.
First impression: the UI is minimal. Left sidebar has Campaigns, Tools, Billing, Settings. No audience manager, no reports tab. Campaign/ad group/ad hierarchy is there but all on one screen. Billing is actually more built out than the campaign tooling, which says something.
When creating a campaign, you pick an objective: Reach, Clicks, or Conversions (coming soon). So for now it's CPM or CPC. Location targeting is hard-capped to US, Canada, Australia, New Zealand. That's it. No UK, EU, or APAC outside ANZ.
Budget setup is standard - campaign total or daily cap. No minimum spend floor, which is a massive change from the rumored $50K pilot entry.
The ad group page is where it gets interesting. You set a max CPC bid (default placeholder is $3.00, which is high) and then there's this thing called "context hints". It's a blank text box where you describe conversations where your ad might be relevant. OpenAI says it's not keywords or exact-match rules - it's soft guidance for the model. No suggestions, no volume estimates. You're writing into a void and hoping the model gets it.
I landed on something like "taking notes during long meetings" for my voice transcription app. Sweet spot seems to be scenario + intent rather than just broad terms. But no one knows best practices yet - the beta only launched two weeks back.
Ad creative is tight: headline 50 chars, description 100 chars, plus a square image. The preview truncates headline at 35 chars, so frontload your value prop. No CTA button selection or A/B testing yet.
Conversion tracking exists but only for web right now - iOS and Android sources are marked "Coming soon". They give you a pixel to install separately, no option to use GA4 or Meta pixels as fallback.
Someone asked about context hints - they're per ad group, not per campaign or creative. And they're about helping the model understand targeting, not just describing your offer.
Another question about whether this feels more like discovery or search - I'd say it's a hybrid. It's not keyword-based intent like Google, but it's not pure interest-based feed like Meta either. It's context-driven placement inside an active conversation. Feels genuinely different.
I've got a campaign submitted and waiting for review. Will report back with real data once it's live - impressions, actual CPC, which context hints pulled volume. If anyone else is in the beta, would love to compare notes on what's getting approved first time and what gets flagged.