在人工智慧领域「红色警报」式的竞争中，OpenAI 预计将于下周日推出一支新的 Super Bowl 商业广告，这将是其连续第二年在 National Football League 最大的赛事中投放广告；但一个问题是：有多少人还记得上一季 Super Bowl 播出的那支 60 秒（=60 s，约 1 分钟）ChatGPT 广告？作者坦言自己几乎完全忘记 OpenAI 的「Intelligence Age」活动，直到在 YouTube 翻出该片段才想起。随著企业销售、消费者订阅，乃至「眼球」争夺愈加激烈，科技公司正把大量资金投入 AI 软体行销，然而一堆 AI 广告往往不是卖出迟钝的使用情境，就是夸大可用功能，反而凸显品牌要展示真正惊艳能力有多困难。
广告衡量公司 iSpot 估计，OpenAI、Google、Meta、Microsoft 这四大玩家在 2025 年于美国线性电视上推广其 AI 服务合计花费 3.828 亿美元（原文 382.8 million 美元），较前一年上升 40%，其中 OpenAI 的主要对手 Google 约占总支出的「一半」左右。与此同时，NBC 表示单支 30 秒（=30 s）Super Bowl 广告的平均成本将达 800 万美元（原文 $8 million）。这种「昂贵讯息」的难题并非新鲜事：Super Bowl 长期以来都是重置叙事、触及巨量新受众的机会，而 Apple 的经典「1984」常被视为科技业（甚至所有产业）的金标准；但人们也容易忘记 Steve Jobs 当年面对的广告竞争有多激烈，许多厂商与零售商以生产力软体、无线键盘、易用性、触控萤幕等各种方式展示 PC 的新意，最终却是那支几乎不展示电脑本体的「奥威尔式」Mac 广告胜出——这种魔法难以复制。
当 OpenAI 在一年前首次于 Super Bowl 亮相时，其行销主管表示目标是赞颂创造力，并与「数千万」对 AI 毫无概念的人产生共鸣；该广告最初以 Sora 影片生成器构思、最终由人类艺术家完成，重点不在导流注册，而在向主流受众传达 AI 只是推动世界前进的工具，并以 ChatGPT App 的黑点脉动动画串起人类进展，最后问观众「你接下来想创造什么？」但反应褒贬不一：Sensor Tower 的资料显示，在 Super Bowl LIX 之后，Google 的 Gemini 更温情的广告短期表现较佳，Gemini 的美国 App 下载量在隔日上升 15%，而 ChatGPT 只上升 3%。不过，Sensor Tower 也指出该月整体而言 ChatGPT 的美国下载增速高于 Gemini，这在其本已领先的情况下尤其值得注意；相比之下，Microsoft 于 2024 年 2 月在 Super Bowl 投放 AI 广告后，Copilot 在当天的美国下载量暴增 342%，其原因可能也反映其旗舰 AI App 先前有多不受欢迎——而该广告面向的观众规模约为 1.23×10^8 名（原文 123 million viewers）。文章最后指出科技公司的一大难题，是如何让「天生是机器」的产品传达人性；AI 尤其令人担忧的是把人类情感与经验外包给聊天机器人，这也使作者更欣赏 Anthropic 以真实生活为根的广告与 Claude 的节制定位；反观 OpenAI 在 Super Bowl 之外的一些广告，常在展示人类片段后以满屏 ChatGPT 文字收尾，让人更想转台，而非更想使用。
Amid code-red competition in AI, OpenAI is expected to unveil a new Super Bowl commercial next Sunday, marking its second straight year advertising in the National Football League’s biggest game; yet the question is how many people even remember last season’s 60-second (=60 s, about 1 minute) ChatGPT spot. The writer admits they had completely forgotten OpenAI’s “Intelligence Age” campaign until digging up the ad on YouTube. As the fight intensifies for enterprise sales, consumer subscriptions, and even “eyeballs,” tech companies are pouring money into marketing AI software, but the pileup of AI commercials—often pitching tone-deaf use cases or exaggerating available features—shows how hard it is for brands to display their most impressive capabilities.
Ad measurement firm iSpot estimates that OpenAI, Google, Meta, and Microsoft collectively spent US$382.8 million (original figure: 382.8 million dollars) promoting AI services on US linear television in 2025, up 40% from the prior year, with Google—OpenAI’s chief rival—accounting for roughly half of the total. At the same time, NBC says a single 30-second (=30 s) Super Bowl ad will average US$8 million (original figure: $8 million). This “expensive messaging” problem is not new: the Super Bowl has long offered a chance to reset the conversation and reach a massive new audience, and Apple’s iconic “1984” is often cited as the tech industry’s gold standard; but Steve Jobs also faced heavy ad competition from many manufacturers and retailers showcasing novel PCs via productivity software, wireless keyboards, ease of use, or touch screens, and the Orwellian Mac ad that barely showed the computer still won the day—magic that is hard to replicate.
When OpenAI debuted its first Super Bowl commercial a year ago, its marketing chief said the aim was to celebrate ingenuity and resonate with “tens of millions” unfamiliar with AI; the ad was conceived on Sora but ultimately crafted by human artists, focusing less on signups and more on telling mainstream audiences that AI is simply a tool that helps move the world forward, ending with “What do you want to create next?” Reaction was mixed: Sensor Tower data indicates that after Super Bowl LIX, Google’s warmer Gemini ad performed better in the immediate aftermath, with US Gemini downloads up 15% the next day versus a 3% uptick for ChatGPT. Still, Sensor Tower notes ChatGPT’s US downloads grew faster than Gemini’s over that month despite ChatGPT already leading, while Microsoft’s February 2024 Super Bowl ad drove a 342% same-day surge for Copilot—perhaps reflecting how unpopular it had been—before a Super Bowl audience of about 1.23×10^8 viewers (original figure: 123 million viewers). The underlying challenge is humanizing a product that is inherently a machine, and with AI there is the added risk of outsourcing emotion and experience to a chatbot, which is why the writer favors Anthropic’s real-life-grounded campaigns and Claude’s restrained framing; by contrast, some OpenAI ads end with a wall of scrolling ChatGPT text that makes the viewer want to change the channel.