Altman has stated publicly on multiple occasions that he is uneasy about the inclusion of advertisements in AI responses, which he considers a "last resort" for businesses.
But faced with reality, Ultraman chose to bow his head.
In the latest Android beta version 1.2025.329 of ChatGPT, ad-related strings such as "ads feature", "search ad", and "bazaar content" were clearly found.
Although these code snippets are only a dozen lines long, they are enough to demonstrate that OpenAI is making technical preparations for embedding ads in ChatGPT.
Admittedly, advertising will generate huge revenue for OpenAI.
However, this would completely destroy the structured, ad-free, and clean image that AI search has built up over the years. Finding a balance between advertising and high-quality content output has become a common challenge for all large-scale search engine companies in 2026.
01
Advertising is poised to become one of OpenAI’s biggest revenue streams.
According to data disclosed by OpenAI, the company spent $2.5 billion in the first half of 2025. At this rate, the company will burn through $115 billion by 2029.
Even though ChatGPT currently has nearly 900 million weekly active users, only about 5% of them are paying subscribers.
Based on OpenAI's pricing, ChatGPT Plus costs $20 per month and ChatGPT Pro costs $200 per month. Even with a 5% paid subscription rate, these subscription revenues are far from enough to cover costs.
To fill this financial hole, OpenAI has developed a monetization plan for free users. According to internal company projections, starting in 2026, the average annual revenue per free user will reach $2, and this figure will grow to $15 by 2030.
The most direct way to monetize is through advertising. OpenAI hopes that by 2030, 20% of its total revenue will come from advertising-related income.
In a recent interview, Ultraman also changed his tune, saying that the advertisement, "while uncomfortable, is not entirely unacceptable." This change in attitude is less a shift in ideology and more a compromise with reality.
Based on internal discussions at OpenAI, its advertising format differs fundamentally from traditional search engines. Altman calls it "intent-driven monetization," the core of which is transforming advertising into "conversational recommendations," rather than simply inserting ad links into search results.
This model works like this: when a user asks for "best running shoe recommendations", the AI will provide professional advice while naturally recommending a paid brand.
According to The Information, the recommendations won't be bluntly labeled "This is an advertisement," but rather integrated into the conversation, such as, "Based on your training intensity, Nike's Pegasus series might be a good fit; its cushioning technology effectively protects your knees." OpenAI may monetize through revenue sharing or charging for "priority access to the recommendation logic."
OpenAI internally developed several ad display prototypes to simulate different presentation methods. One approach was to display sponsorship information in the sidebar of the main ChatGPT reply window, explicitly stating "Sponsored Results Included." This design referenced Google Search's ad display method but attempted to make the ad look more like a "supplementary suggestion" rather than a traditional page ad.
Another approach from OpenAI is to display an ad on the side of the page when the user clicks for further information.
For example, when a user asks for travel advice for Barcelona, ChatGPT will first recommend attractions such as the Sagrada Familia, which are not sponsored content. However, when the user clicks on the Sagrada Familia's information link, they will see a pop-up window containing links to several paid travel service providers.
The purpose of this "secondary trigger" mechanism is to avoid putting pressure on users during the initial conversation.
OpenAI is also exploring a "generative advertising" model. Instead of simply displaying copy provided by advertisers, AI automatically generates customized recommendations based on the user's specific needs and the context of the conversation.
For example, when recommending the same running shoes, marathon runners would be advised to focus on durability and support, while beginners would be advised to focus on comfort and value for money. OpenAI believes that this dynamically generated ad content can theoretically lead to higher conversion rates.
When OpenAI employees were evaluating these proposals, their core point of contention was "how to display ads without alienating users."
At the same time, the team reached a consensus that advertisements should never interrupt or interfere with the natural rhythm of the conversation, and all ad exposures should be set to be triggered after the conversation has progressed to a certain depth.
This principle means that free users will not see ads in every conversation; ads will only appear when the conversation explicitly points to a consumption, travel, or product decision.
02
In November 2025, OpenAI launched the Shopping Research feature, turning ChatGPT into a personal shopping assistant. Users can describe their shopping needs by voice or text, and the AI will ask targeted questions and then provide complete purchase suggestions, including product links and price comparisons.
When OpenAI launched this feature, it explicitly stated that it would not charge any commissions or affiliate marketing fees, and merchants could not pay to influence rankings. However, industry insiders generally believe this is only a temporary strategy. Shopping Research uses a specially trained GPT-5 mini model capable of reading product pages, specification sheets, and credible reviews; this infrastructure can seamlessly integrate with advertising systems.
OpenAI has already partnered with retailers such as Walmart, Target, and Etsy, allowing users to search for and purchase products from these platforms within ChatGPT. While this traffic is currently free, these partnerships will immediately translate into paid promotional channels once the advertising system is live.
During Black Friday 2025, ChatGPT brought a 28% year-over-year increase in e-commerce website traffic, while Cyber Monday saw a 670% increase in retail website traffic driven by AI. If OpenAI can insert ads into this traffic, even with a conversion rate only half that of traditional e-commerce, the revenue will be substantial.
However, in order to counter its impact on shopping portals and advertising revenue, Amazon blocked AI crawlers from several companies, including ChatGPT, Meta, and Google, in the lead-up to Christmas, preventing them from scraping product information.
Meanwhile, Amazon launched its self-developed AI shopping assistant, Rufus, to safeguard its platform entry point and cope with the changes brought about by AI search.
To further explore its advertising business, OpenAI poached Shivakumar Venkataraman, an advertising executive from Google, who had previously led several core projects for Google Search Advertising.
In addition, OpenAI has posted several advertising-related jobs on LinkedIn, including advertising engineer, advertising product manager, and advertising strategy analyst.
These actions all indicate that for OpenAI, the question is no longer whether to do advertising, but how to do it.
Besides cost pressures, competitive pressure also drove OpenAI to make this decision. Google announced in 2025 that it would introduce advertising into Gemini, with a planned official launch in 2026. This poses a direct threat to OpenAI. If competitors are reducing user costs and increasing revenue through advertising, OpenAI will find it difficult to remain unaffected.
Perplexity AI has launched a "Sponsored Answers" feature, clearly indicating sponsored content in search results. While this feature has sparked some user dissatisfaction, it also demonstrates the viability of AI search advertising. OpenAI clearly doesn't want to fall behind in this race.
In terms of market size, the global digital advertising market generates over $1 trillion annually, with Google and Meta accounting for the majority of the share. If OpenAI can enter this market through ChatGPT, even with just a 5% share, it would generate $50 billion in revenue annually, enough to support the company's long-term operations.
However, OpenAI also faces the challenge of gaining user trust. Once commercial interests are incorporated into AI's responses, how can users determine whether a suggestion is based on objective analysis or paid advertising? When ChatGPT recommends a product, will users suspect it's because the product is genuinely good, or because the brand has paid for it?
There is no answer to this question.
OpenAI is currently trying to walk a tightrope between "intellectual neutrality" and "financial self-rescue." If this balance fails, ChatGPT may turn from an "AI assistant" into a "chatbot that sells products."
03
However, it is undeniable that AI search is eroding traditional keyword search.
The market share of traditional search engines is dwindling at a visible rate. According to HigherVisibility's 2025 Search Behavior Study, Google's share in general information search fell from 73% in February 2025 to 66.9% in August, a drop of 6.1% in six months.
The changes in the Chinese market are even more dramatic. Baidu's mobile search market share dropped from 94.72% in 2021 to 58.6% in the third quarter of 2025.
According to QuestMobile data, Baidu's media position index has fallen to ninth place, ranking behind Douyin, Taobao, WeChat, Kuaishou, and Xiaohongshu. Furthermore, Baidu's online marketing revenue in the third quarter of 2025 plummeted by 18% year-on-year, marking its sixth consecutive quarter of decline.
However, these users did not stop searching; instead, they shifted their search behavior to AI tools.
According to QuestMobile's "2025 China AI Terminal Ecosystem Development Research Report," as of October 2025, the number of AI mobile users has reached 720 million, accounting for 50% of China's total internet users. AI search requests exceed 2.8 billion per day, with commercial queries accounting for 41%.
A survey of 1,000 users by the German digital industry association Bitkom revealed that 50% of respondents sometimes use AI chatbots to find information instead of traditional search engines. Among those aged 16-29, 5% rely entirely on AI search, 11% primarily use AI, and 20% use AI at a similar rate as traditional search engines.
In other words, more than one-third of young people have already adopted AI as their primary or sole means of obtaining information.
Even more telling is the change in zero-click rate. Zero clicks mean that after a user searches for keywords, they don't jump to any other page, but instead directly search again or close the page.
SparkToro data shows that Google search has a zero-click rate of 58.5%, reaching a staggering 77.2% on mobile. This means that more than half of searches no longer generate clicks; users leave after seeing the search results and summaries. The core advertising model of traditional search engines is becoming ineffective.
AI-powered search leads to higher conversion rates. Large-scale model search conversion rate refers to the percentage of users whose needs are directly met from the time they initiate a search. ChatGPT's conversion rate is six times that of Google search, indicating that users are more willing to trust the direct answers provided by AI.
Behind this transformation is that AI search has solved the core pain points that have long existed in traditional search.
Traditional search engines only provide a list of links, requiring users to open dozens of web pages one by one to view them. The information quality varies greatly from website to website, and the information is often contradictory, requiring users to verify its authenticity themselves. A simple question often requires a lot of time to piece together a complete answer, and the cycle of "search → click → read → go back → search again" is extremely frustrating for users.
AI search provides complete answers directly, eliminating the need to sift through links, click through them one by one, and piece together information. According to a survey by OpenAI and Harvard University, "How People Use ChatGPT," nearly 35% of users cited "getting direct answers" as the primary reason for using AI search instead of Google search.
AI search supports multi-turn dialogues, which further enriches search results. If the initial question is unclear, users can immediately ask follow-up questions, add conditions, and adjust their direction, gradually clarifying their needs as if conversing with a real person, rather than repeatedly modifying keywords and re-searching. AI can remember the context and form a coherent dialogue, while traditional search treats users as "new users" every time.
The "AI Search User Behavior Research" shows that 72% of AI search sessions involve more than two rounds of dialogue, because this interaction method is more in line with natural human thinking habits. In contrast, traditional keyword searches do not have the concept of "continuous" interaction; each search only contains the current keyword.
Users value the fact that AI-generated answers are not driven by commercial interests and are free from the hassle of pay-per-click ranking. The traditional search engine advertising model has long been criticized; the first few pages of Baidu search results are dominated by paid search ads, drowning out genuine information, and users have to scroll through several screens to find non-advertising content.
While AI can also make mistakes, at least it's not a system where "whoever pays gets priority."
The convenience of voice interaction is also an important factor. Typing on mobile devices is slow, prone to errors, and provides a poor experience when inputting long questions. Voice input, on the other hand, is 3-4 times faster than typing, making it particularly suitable for expressing complex questions. Geokeji's research shows that 58% of users in AI search applications use voice input, while in traditional keyword searches, even with voice input functionality, the usage rate is still less than 20%.
The average question length for voice search is 2.3 times that of text search. Because voice input is more convenient, users can express their needs more completely, improving search accuracy. In scenarios such as driving, cooking, caring for children, and exercising, when hands are occupied, voice search becomes the only option. These high-frequency scenarios have fostered a new habit among users of "speaking directly when needed."
Gartner predicts that global search engine traffic will plummet by 25% by 2026. This is not a simple shift in market share, but a revolution in how people interact with search engines.
From "typing keywords" to "voice describing the complete question", from "filtering a dozen links" to "getting a direct answer", from "single query" to "multi-turn dialogue", the "keyword matching + bidding ranking" model on which traditional search engines rely for survival is being rapidly replaced by the "voice dialogue + direct answer" model based on a large language model.
Therefore, traditional search companies, like ChatGPT, are caught in a dilemma. Adapting to new interaction methods inevitably disrupts their previously effective business models; but staying in their comfort zones means those zones are visibly shrinking. In this transformative period of the upcoming technological revolution, they can only contradict themselves, wavering between two options, yet simultaneously walking a tightrope with ChatGPT on the other side, equally bewildered.
This article is from the WeChat public account "Facing AI" , author: Miao Zheng, and published with authorization from 36Kr.
