Will you trust ChatGPT once it has ads?

• Lots of tech leaders like to say they hate ads.
• Then they decide that, actually, they don't hate ads, because ads can help them make money.
• It seems like OpenAI and Sam Altman are headed that way with ChatGPT.

ChatGPT turned three years old the other day, which means we've spent three years in an AI frenzy.

It also means hundreds of millions of people have been using ChatGPT for years and … not seeing any ads at all, whether they're using the paid version or the free one.

That's not totally astonishing: We've gotten used to consumer internet products like Google, Facebook, and Instagram taking off without ads for a few years. And then, the deluge.

So how much longer will ChatGPT remain ad-free? And what happens when it isn't?

Over the weekend, we got a hint that an ad push may be underway, via some code from ChatGPT's Android app unearthed by developer Tibor Blaho:

Scouring apps for yet-to-be-released features is a long-standing tech hobby, and sometimes it really does yield results. It's also entirely possible that what Blaho found is … something other than an ad product road map.

But it still seems very, very likely that ChatGPT will have ads at some point.

We know this in part because OpenAI executives, starting with Sam Altman, have suggested they will be coming (in 2024, Altman said ads were gross; this year, he allowed that maybe OpenAI could make "some cool ad product").

We know it because OpenAI has been stocking itself with talent from Meta — one of the most successful advertising companies in the world.

We know it because it's simply logical: Altman says ChatGPT has around 800 million weekly users, and only a small percentage of them pay. At some point, his company will want to convert those non-paying users into revenue-generating ones, and ads are the obvious way to do that.

And we know it because on Monday, via an internal memo, OpenAI acknowledged that it was working on ads — but was now delaying that push, while it scrambled to improve its core product in the face of competition from Google. (I've asked OpenAI for comment; while we're here, I'll note that OpenAI has a business partnership with Axel Springer, which owns Business Insider.)

Meanwhile, The Information reports, OpenAI focus groups show that some ChatGPT users already assume ads play a role in the results they're seeing.

But none of that tells us when ads will actually arrive in ChatGPT. And it certainly doesn't address the core question about what happens when you inject ads into an answer machine: Does that machine give you the best answers? Or the answers someone has paid to give you?

Which brings us to the next question: If ads do show up, what would they look like? Because sticking ads into an AI assistant isn't like putting them next to search results or inside a news feed. There's no feed. There's just the answer.

There are a few obvious possibilities, none of which are mutually exclusive:

• Search-style intent ads

This is the Google model: You tell Google exactly what you're looking for — "dumpling spot near me," "best Chromebook" — and advertisers bid to appear next to those queries. If ChatGPT is now a legitimate Google rival, why not use Google's business model, too?

• Personalized ads based on everything ChatGPT knows about you

This is the Meta model: Instead of bidding on queries, advertisers target people, based on what it has learned about their behavior, on and off Meta's properties.

• Old-school text links

The simplest version: "You asked for the best toaster, here are three recommendations, one of which is sponsored." That's basically affiliate marketing. It's low-key and probably the least lucrative.

• Multimedia ads

You are probably typing things into ChatGPT and reading its results. But it doesn't have to work like that: ChatGPT can already talk and show you images. Via Sora, it can show you video. The magic device famed designer Jony Ive is building for the service likely won't have any screen at all. All of which means that Altman and Co. may have a choice to serve you ads that aren't tiny boxes of text on your phone.

But no matter what route OpenAI takes, all of its ad plans will have the same peril: the possibility that injecting paid messages in a service you count on could change your relationship with that service, and weaken that trust.

It's a gnarly problem, even for a company that's used to moving quickly and fixing messes after the fact. They might move more slowly on this one than some people think.