• by Cory Treffiletti , Featured Contributor, Yesterday

I've been a relatively consistent Claude power-user for a while now.  I like Anthropic's approach more than OpenAI's, but I recently tested ChatGPT to see how well it could knock out some design work.

I was surprised. The output was genuinely strong, on par with Gemini and, in a few cases, better, which is not what I expected. As a result, I've been back in there more than I have in the past. In doing so, I ran into the ads.

The ads are basic. I see a sponsored card with a small image, tucked below the answer and clearly marked as separate from the results.  I appreciate that separation.

I have only seen the same advertiser a few times, which I assume is because I'm only doing one narrow thing in that tool right now and the targeting has nothing else to go on.

Nothing about the experience has made me want to close the tab. If anything, I simply notice the ads and get back to what I am doing.  I can’t tell you what the ad was for, so I guess we can argue whether the ad is effective at all, or not.

If you read what others are thinking when they see the ads, you’d think I was insane.

When OpenAI began testing ads for users on both free and paid versions, it met almost instantaneous backlash. Reddit users described the whole thing as a sign of the platform's decline (of course, you have to take most anything you read on Reddit with a grain of salt). 

One widely shared post warned that OpenAI could lose its entire user base after a screenshot showed ChatGPT recommending Peloton in a conversation that had nothing to do with fitness.  That is another overreaction, and the internet is full of them.

I love seeing all these people who complain about untargeted ads and have zero concept of what it takes to build an ads platform, especially one on a consumer-facing site the size of ChatGPT.  As usual, the internet shows its complete lack of patience, understanding, or support.

After I’d written the paragraphs above, I stumbled onto something I found ironic. A lot of that outrage does exactly what outrage always does online: it generates attention. A screenshot of a bad ChatGPT suggestion is a guaranteed few thousand impressions because people love to kick something.

That's not a knock on the people who are genuinely annoyed, because some of them clearly are. It’s just to note a decent chunk of the loudest reactions come from people who understand, consciously or not, that being mad about ads online is a pretty reliable way to generate impressions, which generate ad revenue.  You are yelling about bad ads simply so you can sell more ads. The outrage about the ad model is itself content, competing for the same attention the ad was trying to capture in the first place.

I have been watching the state of ads in chatbots and LLMs, and I think we have two camps: the ads to users and the ads to bots.  I wrote a couple weeks back about the ads to bots, and I’m still mixed on those.  The ads to users will be refined, and here to stay.  That’s an inevitable fact, and some people will complain, but most people won’t care. 

Ask your parents or grandparents who are now using ChatGPT if they care about the ads, and they are most likely going to say, “What ads?”  The format we have now is rather benign and subtle, and simply like paid search ads. 

Maybe that's temporary, and I'll be writing a very different column in six months or a year. For right now, basic and clearly marked is the right amount of ambition for a new ad format inside a chat interface, and I'd rather the industry get that boring part right before it gets too clever.

The louder complaint isn't really about the ad unit anyway. It's about trust, about whether people believe a system that sells placements can still be honest with them the rest of the time. That's a fair question, and it's worth watching closely as this scales.

I'd separate that concern from the actual ad experience I'm having today, which is quiet, clearly labeled, and easy enough to ignore. Most days, that's all I ask of any ad, anywhere.