On May 5, 2026, OpenAI killed the $50,000 minimum spend on ChatGPT's Ads Manager and opened self-serve buying to anyone with a credit card and a $3-to-$5 starting bid. By July it had pushed the pilot into the UK, Mexico, Brazil, Japan, and South Korea. Retail and grocery brands filled more than 100 of the first sponsored placements. The company's own line on all of this is that early data shows "no impact on consumer trust metrics." That sentence is doing more work than it looks like, and it's worth taking apart before the budget meeting does it for you.

Why your brain won't flag a ChatGPT ad the way it flags a banner

Your brain filters a search results page and a sidebar banner as commercial space before you've consciously decided anything. Decades of exposure built that reflex. A conversational answer doesn't trigger it, because you've spent two years training yourself to read ChatGPT's tone as counsel, not inventory. Yahoo and Publicis Media's advertiser survey found 77% of advertisers feel positive about AI in advertising, versus only 38% of the consumers who are supposed to receive it, and 53% of those consumers didn't even know brands were using AI in the ads they'd already seen. That gap is the entire business model. A YouGov poll adds the harder number: 62% of Americans don't trust AI to make ethical decisions, 55% don't trust it to be unbiased. People are being asked to accept commercial influence from the one interface they were told to trust precisely because it wasn't selling them anything.

What "no impact on trust" is actually measuring

It's measuring dismissal rates and click behavior in a pilot where most users still don't expect an ad to appear at all, a baseline that will not hold once sponsored answers are common enough to notice. A 2025 academic study on AI-generated content published in the Journal of Advertising, "Disclaimer! This Content Is AI-Generated," and separate research on what's now called the transparency dilemma both land on the same uncomfortable finding: labeling AI content as AI-generated or sponsored does not reliably build trust. In a meaningful share of cases, the disclosure itself is what triggers the suspicion. OpenAI is optimizing the label to be legally clean. It has not shown the label is psychologically neutral, and those are different problems with different failure modes.

An ad a user doesn't register as an ad isn't performing better. It's borrowing against a trust balance nobody has priced yet.

The budget math that makes this dangerous, not just interesting

This is landing in the worst possible season to get it wrong. 42% of marketers now expect their 2026 budgets to shrink, nearly double the 22% who said the same last year, and retail media (the channel everyone was already rotating into) is growing 17.8% year-over-year toward $69.33 billion in the US alone. A $3-to-$5 CPC with no minimum spend looks irresistible against that backdrop. Compare it to what AI automation is already proving on the platforms that disclose less: Meta's Advantage+ campaigns are running at roughly $60 billion in annualized revenue with a reported $4.52 return per dollar spent, 22% ahead of manually built campaigns. That number is real, but it was earned inside a feed where users already price in advertising. ChatGPT's price only holds while its answers still read as neutral. The day dismissal rates rise and relevance scores fall, which is exactly what the disclosure research predicts, the CPC that looked cheap gets expensive fast, because the auction repriced on a trust premium that just evaporated.

What this means for your next quarter

Test this channel with money you can afford to lose, not budget reallocated from a channel with proven attribution. Track dismissal rate and unprompted brand recall, not just CPC. CPC will look great right up until the moment it doesn't. Stay out of categories where a sponsor tag reads as a betrayal (health, finance, parenting) until independent, third-party trust data exists, not vendor-reported metrics. And treat OpenAI's own numbers the way you'd treat a platform's first-party attribution claims on any channel: directionally useful, not a substitute for your own measurement. The brands that win this cycle won't be the ones who bought earliest. They'll be the ones who noticed the trust premium before the auction did.