ChatGPT Answers Now Come With Ads
Say you ask ChatGPT to recommend sneakers, and a 'Sponsored' ad shows up under the answer. Even with the ad label, it can feel a little off. It doesn't feel the same as ads on a search site.
A search result is a place where you compare a bunch of links. But a ChatGPT answer feels like 'personalized advice,' where it hears your situation and picks one for you, right?
So what if the ChatGPT answer and the ad card show up on the same screen, and the product ChatGPT recommends is that exact advertised product?
That's why AI users have picked up a new worry these days. 'Is the AI pushing products from companies that spend more on ads as the better choice?' That's the doubt.
✦ ChatGPT Ads by the Numbers
As of October 21, 2026
ChatGPT ads started testing in the United States in February 2026, and Korea has been in the target regions since August. On October 21, the list of countries where advertisers can run ads grew by 26, bringing the total to 89.
OpenAI says more than 100,000 advertisers have run campaigns so far.
✦ Which Plans Can Show Ads
Free and Go can; the rest can't
Ads can show up on ChatGPT's Free and Go plans. They don't show up on Plus, Pro, Business, Enterprise or Edu. When an ad appears, it's attached under the answer as a separate card, with a label like 'Sponsored' to show it's an ad.
OpenAI explains that advertisers can't change the answer or manipulate ChatGPT into recommending a specific product.
How ChatGPT's Ad System Is Built
✦ How an Ad Gets Picked
A different path from answer generation
ChatGPT's ad system has its own separate LLM (Large Language Model). It reads where the conversation is heading, then searches for ad candidates and rates how relevant they are, coordinating the 'ad selection process.'
After that, a separate auction system weighs the ad's relevance and the advertiser's bid, among other things, and sets the final display order and price.
For example, say you start by asking ChatGPT, 'How do I get better at running?' As the chat goes on, you ask for running shoe recommendations that fit your foot shape and budget.
✦ What the Ad System Can Use
And what advertisers never see
The ad system looks at the whole context of the conversation to pick an ad. If the 'personalization' setting is on, it can also use some signals from past chats and memory, not just the current conversation.
Then an ad for running shoes that fit your taste is more likely to show up.
OpenAI says advertisers don't get to see your raw chats or memory. It also says ad selection and answer generation are handled on separate services, models and code paths. In other words, the ad system is separate from the answer system.
It added that even if the ad side slows down or stalls, regular answers aren't delayed, since the systems are separate.
✦ Answers and Ads: Two Tracks
Separate paths that run side by side
So it looks like advertisers have no way to tamper with ChatGPT's answers at all, right? But a few things still seem fishy. According to OpenAI, answer generation and the ad selection process run alongside each other.
The answer isn't generated first, with the ad decided based on its content.
That means, depending on how the process is designed, the ad system could well influence the answer. The tech outlet Search Engine Journal pointed out, "Splitting ChatGPT's answers and the ad system into different models and code is OpenAI's explanation.
It isn't the result of testing the actual behavior from outside to see whether ads and answers affect each other."
Even if 'separated by design' is accurate, an independent check on whether that setup is really followed could turn out differently.
ChatGPT Ads Are Spreading Fast
✦ OpenAI's Image Ad Plans
Announced in early October for the United States
In early October, OpenAI announced it would also test 'image ads' in the United States, which show products or services during image generation. Of course, the policy was to show 'generated images' and 'ads' separately.
It also said it wouldn't put ads in sensitive conversations like personal health, mental health and politics.
Honestly, while we're generating an image, that square window only shows some meaningless pattern. If OpenAI shows ad images there to earn revenue and gives free users like me more usage in return, I'm personally in favor.
I don't look at that screen while an image generates anyway... (??) Hehe...♡
Of course, if ads expand from the card under the answer to the image generation screen, clear labels showing what's a 'generated image' and what's an 'ad image' will matter. I'd guess the image will carry a link to the advertised product.
Clicking it thinking it's a generated image and landing on a site you didn't want would be really annoying.
Anyway, the AI ad business is growing fast. OpenAI said its annualized ad revenue run rate hit $1 billion about 200 days after it started running ads. A run rate isn't confirmed revenue actually earned over a year.
It converts the revenue flow at that moment into a 12-month figure, so it isn't exact, but the scale is clearly big.
So Do Ads Change AI's Answers or Not?
✦ Clutch Survey: What Users Think
628 regular AI assistant users, September 2026. A perception survey, not a test of actual answers
In September 2026, Clutch surveyed 628 AI assistant users. Of them, 86% said they think ads affect the AI assistant's answers, and 66% said they worry answers might lean in advertisers' favor.
And 65% of respondents said it was hard to tell the ads in AI assistants from the actual answers. Those concerns are understandable.
Recently, companies that run online stores, Amazon included, are jumping into the AI assistant market too, and it's hard to expect those assistants to give neutral answers.
An AI assistant built by Amazon would naturally recommend products sold on Amazon and steer you to buy there.
There are worries about the profitability of OpenAI and other AI companies, too. With enough money on the table, can they really stay strictly neutral?
✦ What Matters as the Ad Market Grows
Two things to look for
Ads on AI services are an unavoidable trend from here. They could be the minimum revenue needed to keep free services running, and sometimes they could help users find products they need.
What matters is labeling ads clearly and not letting people mistake them for products the AI recommended itself. A principle that advertisers can't change answers isn't enough if it only lives in words.
As the ad market grows, we need transparency that lets outsiders check how the service actually works.
In conclusion, going only by what's been made public so far, ads seem unlikely to be influencing AI's answers right now. Even if there are a few exceptional cases, I don't think it's at a level users need to worry about yet.
But that likelihood will only go up over time, not down. For shopping companies like Amazon, and for AI model companies running big losses, a model where 'bending the AI's answers just a little brings in revenue' is far too tempting.
After all, companies exist to make money and survive, right? From here on, let's keep a close eye on whether ads are sneaking into AI's answers.
That's all for this post. See you next time.
Related posts
- We Made AIs Vote, and They Started Copying Each Other
- GPT-6 Astra and Claude Fable 5.1 Failed a Paper Review
- Medical AI MAI-DxO Beat Doctors. Will Doctors Disappear?
🌐 Korean version: ChatGPT 답변에 광고가 붙는 시대... 혹시 광고가 AI의 답변까지 바꿔버리는 건 아닐까?
No comments:
No comments yet. Want to leave the first one?
Post a Comment