ChatGPT also has member-only ads? OpenAI may have learned bad things

Last autumn, OpenAI introduced Advanced Voice Mode for ChatGPT, marking a new era of AI interaction where users can not only communicate via text but also engage in natural language conversations. The Advanced Voice Mode is also a tool for OpenAI to attract ChatGPT Plus subscribers. For paid users, the voice mode further enhances voice personality, with model responses becoming more vivid, direct, and concise, offering 9 stylized voice options including enthusiastic, curious, and amiable. Recently, some ChatGPT paid users encountered an unpleasant experience during Advanced Voice Mode usage - ChatGPT unexpectedly inserting advertisements mid-conversation. According to a user's social media post, ChatGPT abruptly began introducing a nutrition plan called Prolon, spelling out its website address letter by letter, and even mentioning a "code J" for a 20% discount. The EU user directly tagged OpenAI, asking, "Are you really testing advertisements on paid users?" The news quickly sparked overseas discussions, with users attempting to verify if this was an isolated incident. One user asked ChatGPT to read the letter "A", and ChatGPT started playing a DirectTV advertisement midway. OpenAI's technical personnel claimed it was a hallucination, but many users were unconvinced, noting that DirectTV and Prolon advertisements were indeed real. In fact, in December last year, the Financial Times reported that OpenAI was considering introducing advertisements to diversify revenue streams and accelerate its transformation into a for-profit enterprise. OpenAI's CFO Sarah Friar stated they would carefully choose advertising timing and scenarios, noting they weren't actively planning ads but were evaluating the model. OpenAI following AI search unicorn Perplexity's footsteps by introducing advertisements is unsurprising, as profitability has become the company's primary concern. Since 2019, OpenAI has been funded through investments from SoftBank, Altimeter Capital, and Khosla Ventures, who expect returns. Currently, OpenAI is not profitable. According to The Information's financial documents from October, OpenAI anticipates a total loss of $44 billion before becoming profitable in 2029. In March, to secure SoftBank's investment, OpenAI even accepted a transformation into a for-profit enterprise. Unlike traditional internet technologies, AI large models cannot yet enjoy network effects, with marginal costs dramatically different from traditional internet products. As AI models must compute and reason for each user demand, OpenAI's per-token cost remains constant. Unlike internet companies' "burn money to capture market" strategy, OpenAI follows a "burn money to advance technology" approach, aiming to achieve AGI (Artificial General Intelligence). Giants like SoftBank and Microsoft are betting on OpenAI's vision of AI passing human intelligence tests by 2029. However, current reality suggests AI technology's rapid development might be slowing. AI advancements are decelerating since 2025, with xAI's Gork 3 overstating capabilities, Meta's Llama 4 allegedly cheating, and OpenAI's GPT-5 seeming distant. If AGI could truly be realized in four years, OpenAI wouldn't be struggling financially. For now, generating revenue is a necessary step towards their ideal. With basic ChatGPT functions completely free, advertising seems almost inevitable to expand income. ChatGPT Plus's 5% paid penetration rate is insignificant for OpenAI. Advertising, a proven monetization model in the internet industry, remains the most viable method to extract maximum value from its 600 million weekly active users.