OpenAI’s advertising pilot is live, but advertisers can’t measure ROI. Here’s why that matters more than the ads themselves.

The headline sounds bigger than the reality

OpenAI is running ads inside ChatGPT. That’s the headline, and it’s generating exactly the kind of excitement you’d expect. A new platform, a massive audience, first-mover advantage — the pitch writes itself.

The detail that matters: advertisers in the pilot have no way to verify whether thespend is working. No conversion tracking. No attribution. No ROI measurement. Justimpressions and a hope that proximity to AI means proximity to intent.

‍

What the pilot actually offers

From what’s been reported, the ChatGPT ads pilot is display-style advertising shown within the ChatGPT interface. Advertisers get placement, but the measurement stack isn’t there yet. There’s no pixel, no click-through attribution model, and no way toconnect an ad impression to a downstream conversion.

This isn’t unusual for a brand-new platform — early YouTube and TikTok ad products had similar gaps. But the marketing press is treating ChatGPT ads as a performance channel, and that framing is dangerous for anyone managing a budget with accountability attached.

‍

Why “no ROI measurement” is the only detail that matters

If you’re a DTC brand spending £50k–£500k a year on media, every pound needs to justify itself — or at least be part of a deliberate brand-building strategy with a clear thesis. ChatGPT ads currently offer neither.The problem isn’t that the ads won’t work. They might. The problem is you can’t know. And for a marketing manager reporting to a founder or CMO, “we think it’s working” isn’t a position you want to be in.

‍

The comparison that should make you pause

Google Ads and Meta Ads — for all their flaws — give you a measurement framework. You can debate attribution models, question data-driven vs last-click, and argue about view-through conversions. But the data exists. You can make informed decisions, even if those decisions involve some uncertainty.

ChatGPT ads, right now, give you none of that. Spending on a platform with zero measurement isn’t bold. It’s guess work with a budget.

‍

When this might change — and what to watch

OpenAI will build measurement tools. Every ad platform does eventually. The signals to watch:

• Pixel or tag integration. When OpenAI launches a tracking pixel or integrates with Google Tag Manager or a CDP, the measurement gap starts closing.
• Conversion API. A server-side tracking option would signal serious intent tocompete for performance budgets.
• Third-party verification. When platforms like DoubleVerify or IAS can verify ChatGPT ad delivery, brand safety concerns ease.
• Case studies with actual numbers. Not “we saw a lift in awareness” — real ROAS or CPA data from advertisers willing to share.

Until at least two of these exist, treating ChatGPT ads as a performance channel ispremature.

‍

What to do now

1. Don’t allocate performance budget. If your spend needs to demonstrate return, this isn’t the place for it yet. Keep it in channels where you can measure.
‍

2. Consider a small brand test — if you can afford to. If you have a separate brand awareness budget with softer KPIs, a small test could be worth running for learning purposes only. Set the expectation internally that this is R&D, not performance.
‍

3. Watch the measurement roadmap. OpenAI will announce tracking capabilities when they’re ready. That’s the moment to reassess, not before.

4. Don’t let FOMO drive spend. The “first-mover advantage” argument assumes the platform matures and early adopters benefit. That’s possible but not guaranteed. Late movers with better data often outperform early moverswho guessed.

‍

FAQ

Is anyone actually running ChatGPT ads?

Yes — the pilot includes a small group of advertisers. But the lack of public performance data suggests results either aren’t impressive or aren’t measurable enough to share.

Could ChatGPT ads work for top-of-funnel?

Potentially. The audience is large and engaged. But even top-of-funnel campaigns benefit from some measurement — brand lift studies, assisted conversions, view-through data. None of that is available yet.

‍

Should I be worried about competitors getting in first?

No. If the platform can’t measure performance, early spend doesn’t create a measurable advantage. When measurement arrives, the playing field resets.

‍

‍

‍