Why I Am Bullish on ChatGPT Ads

I have done SEO for more than twenty years, since Google was the new search engine, and I ran Google Ads campaigns back then too. So I have watched a search company build an ad business once before. Ad-supported ChatGPT is going to work, and I am bullish on it.

The numbers already point that way. OpenAI started testing ads in the US on February 9 for Free and Go users, and by August 31 it reported that ChatGPT Ads had reached $1 billion in annualized revenue in under 200 days, with tens of thousands of advertisers, availability in over 40 countries, and self-serve buying in 31 European markets. But the revenue is not why I am bullish.

In February 2002, Google switched AdWords to cost per click and ranked ads by a combination of click-through rate and the price the advertiser agreed to pay. You picked keywords, you set your bid, and what you paid depended on who else was bidding, what they bid, and how often people clicked each ad. An ad nobody clicked lost its spot no matter how much money was behind it. The ads sat in their own column next to the results, and the press release promised that the organic results were "unaffected by payment". I think that was the best advertising system ever built. It worked for advertisers, and it worked for users. Today everyone hates everything to do with ads, but back then the ads were a feature. They were relevant, I clicked plenty of them myself and found what I was looking for, and you could always choose to look at the results or look at the ads, instead of the commingled mess we have now.

If ChatGPT ads become anything like that, I am all in. On paper, that is what OpenAI is describing. Its stated principles say ads are "always separate and clearly labeled" and "do not influence the answers ChatGPT gives you", the first format is an ad at the bottom of an answer when there is a relevant sponsored product, and Plus and Pro stay ad-free. Free users can even opt out of ads in exchange for fewer daily messages. I do not know what the final format should look like, but a labeled ad below an answer that was written without the ad in mind is the 2002 shape, not the 2026 Google shape.

Meanwhile Google is trying to become ChatGPT. At I/O in May it announced the biggest change to its search box in 25 years, "completely reimagined with AI", with Gemini 3.5 Flash as the default model in AI Mode. When the incumbent starts copying the newcomer, the original usually wins. Most people still use Google because it is what they know. Many of them do not think of ChatGPT as an alternative at all. But think back to using AltaVista or HotBot and then running your first Google search. It was obvious within seconds that Google was better, faster and more accurate, so you switched and never went back.

You want an answer, so you ask Google, because that is what you know. But those answers are not good. People expect Google to be instant, and by Google's own count it handles more than 5 trillion searches a year. If you want an AI answer on that many queries, delivered instantly, you cannot run a strong model. You run the fastest and cheapest one you have, and I think that is why the default in AI Mode is a Flash model. So the answer leaves a bad taste. You read it and think, that cannot be right, and then you find out it was wrong. At some point someone says, why did you not ask ChatGPT, and you try it, and the answer is much better.

That gap did not exist a year ago. Back then, normal ChatGPT web search summarized the top ten results for you, roughly what Google's AI Overview does. The alternative was deep research, which OpenAI launched in February 2025 on a version of o3, at 5 to 30 minutes per query and 100 queries a month on the $200 Pro plan. It was excellent, and people were rightly excited about it, but you used it for serious research, not as a replacement for a web search. A year ago I would ask ChatGPT about a topic I wanted to understand, and I would still go to Google when I needed new running shoes.

That has completely changed. I use a paid account and only the Pro models and features, so I cannot speak for the free version, though since August the free tier has run GPT-5.6 Luna with unlimited text chats which is a rather good model. If I ask ChatGPT a question I would have typed into Google, it scans the web, investigates, and does the thinking for me. It goes deeper than a summary. It is as skeptical as I am, so when it hits SEO spam or an astroturfed Reddit thread, it dismisses it and moves on. It can search, scrape and use websites, so it can find the best running shoe, the best places to travel, or the cheapest price for something. It is still not instant. I wait 10 to 20 seconds for that kind of answer, and either I get used to that or ChatGPT gets faster. But when I want to research anything now, ChatGPT is the place, not Google.

Running shoes, travel destinations and price comparisons are exactly the queries ads work on. In March 2025 I wrote about the chatbot monetization problem, where chatbots got the informational questions while Google kept the profitable commercial ones. That has flipped. The commercial questions are moving to ChatGPT, because ChatGPT answers them better.

The objection I hear is that once ads exist, they will creep into the answer itself, and you will never know whether you are being sold something. I get that, and it is a valid concern. It is also exactly the concern people had about Google from the start and still have. It is a conflict of interest. I am not going to judge how well Google has handled it, or how well OpenAI will. What I can say is that it has not hindered Google's economic success at all, and I do not think it will hinder ChatGPT's either.

What I'm building

Delegate tasks. Get software.

Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.

Take a look at vroni.com