From 'Things to Show' to 'Responding to Desires': The Future of ChatGPT Advertising Through the Lens of the GPT-5.5 Instant Update

OpenAI has announced a new version of GPT-5.5 Instant. The official post explains that "conversations are now even more enjoyable," "it can understand the intent of questions more accurately and adjust responses accordingly," and "it makes shopping and chatting more comfortable." Furthermore, the official ChatGPT account announced that the new GPT-5.5 Instant is being rolled out sequentially to Pro and Plus users, with plans to reach free users by tomorrow.

I believe this is more than just a model update. I sense that OpenAI is attempting to evolve ChatGPT from a "smart question-answering system" into a "life infrastructure that receives human desires and directs them to appropriate solutions." Beyond that, there is the potential to reshape the very history of advertising.

OpenAI has reached a point where monetization is unavoidable

OpenAI has over one billion monthly users (with about 5% being paid subscribers). However, the development of the AI models used by non-paying users and the operation of data centers incur enormous costs. According to reports, while OpenAI is generating significant revenue, its research and development, computing resources, and sales and marketing expenses are also ballooning, leading to massive operating losses.

In other words, rather than continuing as a volunteer service indefinitely, ChatGPT has reached a stage where it must build a sustainable revenue system while serving as a convenient intellectual infrastructure for humanity. The user base and computing resources are too large to be supported solely by paid subscriptions like Plus and Pro. That is precisely why how to monetize the massive user base, including free users, has become an unavoidable theme for OpenAI.

This is where advertising comes in. However, advertising in ChatGPT must not be the same as traditional web advertising. If they get this wrong, instead of gaining a major revenue opportunity, OpenAI will damage the greatest asset ChatGPT has: "trust."

The revenue system ChatGPT should build is not an advertising slot, but desire routing.

Google built a massive revenue engine through search advertising, which supported its non-paying users. Human desires appear in a very concentrated form in the search bar. This is because users input what they want in a somewhat verbalized state, such as "hotel," "insurance," "moving," "English conversation," or "MacBook." By connecting advertisements to that search intent, Google turned ads from noise into, in some cases, a shortcut.

However, what is thrown into ChatGPT exists even further upstream than search terms. Before saying "which product I want," users talk about "wanting to feel this way," "thinking this thing was cute," "having trouble with this recently," or "not knowing which one to choose." What exists there are unorganized desires that have not yet become product names or search terms.

For example, suppose a high school student says, "I saw a girl with a cute smartphone case on the train today; it was pink, had a strawberry pattern, and I really want it." In a traditional e-commerce search, you would just display a list of products using the keywords "pink," "strawberry pattern," and "smartphone case." However, ChatGPT can understand from the context of the conversation what that person usually calls "cute." Is it a pale, translucent Korean-style aesthetic, a Heisei-retro strawberry milk vibe, Sanrio-esque sweetness, Y2K, or a mass-produced style? In other words, it can treat the word "cute" as a coordinate of that individual's aesthetic sense.

This is the essence of ChatGPT advertising. Advertising should not be the entrance, but commercial information that happens to occur as a result of processing a user's desire. If what the user wants can be handled through conversation, then the chat should suffice. If it can be satisfied by image generation, then create an image. If Deep Research is needed, then look it up. If a product is needed, only then does the product appear. In other words, product recommendations should not be "to sell," but the result of responding to a user's desire.

Concerns regarding advertising partnerships with Dentsu and Hakuhodo

In Japan, OpenAI has begun domestic pilot operations for ChatGPT advertising in collaboration with companies like Dentsu Digital. I can understand that the sales networks and coordination capabilities of advertising agencies like Dentsu and Hakuhodo are necessary to connect with advertisers in the Japanese market and to handle legal affairs, screening, brand safety, and business customs.

However, there are also major concerns here. The Japanese advertising industry has a culture of building strategies based on past advertising theories such as AIDMA, AISAS, SIPS, personas, and customer journeys. These were effective auxiliary lines in the media environment of the time. In the mass media era, it was necessary to set up personas as representative consumer images because individual users could not be known. In the web advertising era, it was necessary to estimate what users wanted based on browsing and search history.

However, with ChatGPT, that premise collapses. Even without setting up a persona, the user themselves talks to you every day. You don't need to create a fictional representative like "35-year-old woman, living in Tokyo, high health consciousness"; you can read directly from the conversation what that person is currently struggling with, what they want to make more convenient, and what they want to be comforted by. The unit of advertising shifts not from attributes or personas, but to the state of desire at that moment.

That is precisely why ChatGPT advertising should not be treated as a "new advertising surface." The idea of placing banners in the chat window, inserting products that seem to fit the context, or designing brand contact based on emotions is too old-fashioned. ChatGPT is not a medium; it is a place where users verbalize their own desires. If you bring an old advertising OS into that space, the precious conversational infrastructure will just degrade into a high-precision advertising slot.

Advertising that hacks the reward system should be ended

Advertising until now has often evolved in a direction that dulls user judgment rather than helping it. Banners that stay at the bottom of smartphone screens, full-page ads that are hard to close, interstitials that require watching a video to close, and UIs that induce accidental taps. The same structure applies to Rakuten's massive point returns and shopping rallies, as well as TV shopping tactics like "call now," "first 100 callers only," "this time only," and "get a gift of pickled vegetables from a long-established Kyoto inn."

Of course, this is not to say that all traditional advertising strategies are malicious. However, in those cases, feelings of scarcity, security, value, and social reward are pushed to the forefront rather than the value judgment of the product itself. For example, when selling a set of five lower back pain relief corsets, the question "do I really need that many for myself?" should naturally arise. However, when a line like "you'll make your neighbors happy if you share them" is inserted, excessive purchasing is converted into goodwill. It becomes not overbuying, but thoughtfulness.

Such advertising does not understand the user's desire but stimulates the reward system to drive purchasing. Points are a good deal, it's only for now, someone will be happy, it's safe because it was introduced on TV. These mechanisms do not support the user's judgment; they weaken the brakes on their judgment.

ChatGPT must not move in this direction. ChatGPT should not be a place that rushes a user's judgment, but rather a place that helps them settle their judgment. It will not earn user trust unless it is an AI that can say, 'You don't need to buy this right now,' 'What you already have is enough,' 'This is cute but too expensive,' 'The reviews look suspicious,' or 'Five is too many; you should try one first.'

What AI makes possible is,
not ads for selling, but recommendations that aid in decision-making.

Recommendations themselves are not evil. In fact, in an age where there is too much information, reliable recommendations are necessary. The problem is whether they are recommendations for the advertiser or for the user.

Where ChatGPT truly holds value is in its ability to become a purchasing agent for the user. Amazon and Rakuten have vast inventories, but their search results are tainted by ads, SEO, rankings, and suspected review manipulation. Users don't want to search for products; they want to find the 'right answer for themselves.' That is why there is value in an AI that you talk to every day making recommendations based on your preferences, budget, lifestyle, usage habits, dislikes, and past mistakes.

In this future, Amazon and Rakuten will not disappear. They will remain as providers of actual purchase destinations, inventory, shipping, returns, and price comparisons. However, the place where users search first will be ChatGPT, not e-commerce malls. Users will ask ChatGPT, 'What should I buy?' and ChatGPT will compare options, including multiple sellers, direct-from-manufacturer, used items, physical stores, subscriptions, or even the option of 'not buying anything at all.'

A strategy of increasing exposure because
sponsors paid OpenAI more in advertising fees is a no-go.

The method of increasing exposure because an advertiser paid OpenAI more in advertising fees will not work here. Sponsorship must be clearly disclosed. And if a non-advertised product is truly a better fit for the user, the AI should recommend the non-advertised product. Standing by the user rather than pandering to advertisers is what will ultimately increase advertising value. This is because users will think, 'If ChatGPT recommends it, it's really okay to buy.'

The history of advertising is also a history of where trust resides. There was an era when things were trusted because they were in the newspaper. There was an era when things were trusted because a famous actor recommended them in a TV commercial. On the internet, it was thought that by tracking browsing and search history, one could understand what a person wanted. But in the next stage, the one who truly knows what a person wants will be the AI that listens to their words every day.

A healthy form of advertising in the AI era

Therefore, the ideal form of ChatGPT advertising is not 'showing ads.' A user throws a desire at ChatGPT. If that desire happens to be solved by a product—rather than through conversation, image generation, research, scheduling, or creative support—commercial information appears as a result. I believe this is the healthy form of advertising in the AI era.

OpenAI partnering with the Japanese advertising industry is realistically unavoidable. However, they must not hand over the OS of the advertising experience. They can leave market development, sales, screening, legal affairs, and advertiser connections to agencies like Dentsu and Hakuhodo. But the fundamental design of 'what an ad is within ChatGPT' should remain firmly in OpenAI's hands.

ChatGPT must not become a space filled with ads like the mobile web. Instead of ads that hack the user's reward system and dull their judgment, it should be about recommendations that aid the user's judgment. Ads should not be something that pollutes the conversation, but something that appears only when necessary as a result of processing a desire. The GPT-5.5 Instant update looks like an important step toward moving in that direction.