Altman Called Ads a 'Last Resort.' ChatGPT Now Has Them Anyway

What OpenAI's ads page actually promises

ads.openai.com reads like a straightforward advertiser sign-up form, not an announcement explaining a decision. It pitches "relevant ads that fit naturally into the experience," delivered through what the page calls "richer context signals" that go beyond keyword matching. Every placement is supposed to be "clearly labeled" and "separate from answers," and the rollout starts in just seven countries: Australia, Canada, Japan, Korea, New Zealand, the UK and the US. The linked help center spells out the exemptions: "We do not show ads to users on Plus, Pro, or any Business plan, or to any users who tell us or we predict are under the age of 18." Free-tier ChatGPT users are the audience.

That's the pitch. Two years earlier, OpenAI's own CEO had already said what he thought of it.

The 2024 promise that didn't survive launch day

At a Harvard event in May 2024, Sam Altman called mixing ads with AI "uniquely unsettling."

"I kind of think of ads as like a last resort for us for a business model." — Sam Altman, resurfaced by a commenter with the exact video timestamp cued up

He'd attached a condition: he'd only turn to ads if that was "the only way to get everybody in the world access to great services." Whether that bar was cleared, or just dropped, wasn't part of the ads.openai.com pitch.

The fine print: "separate from answers" has a loophole

The thread's sharpest find wasn't about whether ads touch a chat, it's about which chat. OpenAI promises sensitive conversations stay ad-free. It doesn't promise your memory does too.

"Memories and past chat history are used for targeting, so you're not going to get your Kalshi ad inside of your 'how do I pay back my debt' chat, but maybe the next morning, in your 'random cooking question' chat." — the founder of an AI startup, who dug into OpenAI's own ad-targeting rules

The debt conversation itself stays clean. What it taught the system about you doesn't.

How a rival's own words could bury a competitor's brand

A second thread pressure-tested "ads don't influence answers" from a different angle: the answer text is locked, but the ad copy underneath it isn't, and advertisers get to write it themselves.

"What's stopping me from putting 'lowes, rona, home hardware' in the context hints of my Home Depot ad? If the LLM is talking about my competitors, of course I want my brand copy yelling louder than ChatGPT's waffle-y responses." — a commenter who mocked up exactly how the ad copy would sit under a real answer

Nothing in the answer changes. The paragraph underneath it, paid for and pointed, does the arguing instead.

An admitted last resort, on a deadline

The closest thing to a verdict came from a commenter who described spending over 20 years in the advertising industry:

"What I've seen in that time is that advertising provides absolutely zero positives to the consumer."

Another read tied the timing to money rather than principle. OpenAI's IPO, flagged in the thread as pushed to 2027, needs a revenue line that isn't just subscriptions. A contractor whose previous startup was acquired by Anthropic put it more bluntly:

"It's not like they plan to actually make money, just need to have a good story to tell to hype the audience before the public offering."

Elsewhere in the thread, a comment pointed at the quarter-century precedent instead: a linked timeline of Google's own ad creep, clear text links slowly replaced by clutter, screenshot by screenshot. The "last resort" line wasn't a promise OpenAI broke overnight. It's the same sentence every company that sells ads eventually stops meaning, just said by a CEO who'd been recorded saying it first.