Gemini 将成为美国学生不得不用的 AI。
近日,Google 宣布了一项看似让步的政策调整:从明年初开始,其 K-12 教育账户中的 AI 工具将默认关闭,学校管理员可以自行决定是否为学生启用 Gemini 等能力。消息传出后,不少教育工作者松了口气,觉得终于有了挑选的余地。
但重点藏在了「小字」里,附带条款提到「搜索和 AI 是深度融合的」,不能二选一。

换言之,一旦学校挑选屏蔽 AI,学生将连 Google 搜索也一并失去,显示屏上只会表明一条来自学校的提示,说明该服务暂不可用。
这个时代的教育肯定要拥抱 AI,但被逼着使用一款 AI 的课堂,已经开始变味了。
我画地为牢啊,Google

许多学区管理者陷入了两难。一边是家长和教师对 AI 在校园扩张的忧虑,另一边是放弃搜索引擎的不可承受之重。
Google 在美国 K-12 教育领域的覆盖程度,早已超过了大多数人的日常感知。

Google Workspace for Education 覆盖逾 1.7 亿名学生和教师,Gmail、Docs、Classroom 组成了许多学区无法绕开的操作平台,从布置作业到批改、交流,全在里面。搜索引擎更是一种基础设施,像教室里的白板一样默认存在。
前联邦贸易委员会主席 Lina Khan 认为:Google 正在「利用其搜索垄断地位胁迫学校」接受其 AI 机型。多个法院已裁定 Google 违反反垄断法,而此次针对学校的操作,恰恰印证了这家公司「毫不收敛」。
南加州大学教育学教授 Morgan Polikoff 说,这是科技行业「不惜一切将 AI 强加给所有人」的典型案例,即便对象已明确表示拒绝。
从商业逻辑看,这个捆绑策略并不难理解。教育市面不仅意味着数以千万计的使用者,更意味着习惯的早期塑造。让一代学生从小在 Google 生态中成长,借助 Gemini 完成作业、使用 AI 工具预习和复习,这些行为日后转化为付费使用者的概率,远高于成年后才第一次接触机型的人群。
默认变了
有必要先回顾一下它怎么走到今天这一步。
2025 年之前,Google 在教育领域的 AI 能力基本处于可选状态。Gemini 是可以单独申请访问的附加工具,学区有明确的开关可以控制。
2026 年 8 月,Google 悄悄做了一件事。它为此前已允许学生独立访问 Gemini 的 K-12 学区,默认启动了 Google Classroom 内的 AI 能力。
这个动作激怒了部分学区管理者。他们并不认为自己当初允许学生访问独立的 Gemini 账户,就等于授权了整个 Classroom 平台默认启用 AI 工具。
Google 随后宣布「反转」这一决定,表示将在 2027 年初恢复学区对 AI 的自主控制权。但条件也随之而来——关闭 AI 意味着同时失去 Search。反转附带了代价,代价由学区来承担。
同时,Google 还宣布为允许启动 AI 的学区给出两种搜索模式:一种是与消费者版本相近的感受,包含 AI Overviews 和 AI 对话,另一种是基于 Gemini 的「Guided Learning」学生版,主打循序渐进的引导式学习。

其他搜索引擎对 AI 能力的处理方式并不如此,DuckDuckGo 在设置中给出了完整的 AI 能力开关,使用者可以关闭 Duck.ai 和 Search Assist,或者直接访问 noai.duckduckgo.com 取得无 AI 的搜索感受,官方明确宣传这一点与 Google 不同。

Brave Search 同样给出了关闭「Answer with AI」的设置选项,只需切换一个开关。微软 Bing 虽然也有 AI 摘要能力,但给出了官方浏览器扩展,允许使用者一键切换 AI 能力的开关。

相比之下,Google Search 的 AI Overviews 被官方定性为核心能力,不可关闭。唯一的变通,是在搜索结果页点击「Web」过滤器,或者在搜索时附加 \&udm=14 参数。
孩子的大脑,谁来保护
2026 年 9 月,美国两个最大的学区——纽约市和洛杉矶,相继宣布在学校设备上对生成式 AI 实施广泛限制。
纽约市禁止 8 年级及以下学生使用 AI。洛杉矶统一学区则希望学生可以继续使用 Google Search,但关掉其中的 AI 摘要能力,理由是摘要会让学生不再主动寻找多个信息来源。
纽约州随后借助了一项聊天机器人安全法,明确禁止科技公司向未成年人给出多种 AI 能力,这是美国各州在 AI 教育监管上迈出的较具体的一步。

这两个学区都是 Google 的客户,也恰好是这场争论里最具代表性的声音。
家长群体的反弹也在持续积累,他们对 AI 的顾虑,往往是对显示屏时间更大争议的延伸。AI 只是把这个争议推到了一个新的尖锐程度。

▲ 图源:Penn State Health News
孩子使用 AI 的风险,可以从几个层面来理解。
最直接的是作弊,Gemini 可以从容完成学生的写作作业。Google 为此推出了面向学校的「Guided Learning」版本,声称会限制 AI 直接给出答案,引导学生逐步思考,模拟一对一辅导的感受。但经过测试,即使是消费者版本的 Guided Learning,也能针对一道 AP 历史题生成一篇完整的作文。
学校版本据称有更多限制,实际上不过是「吩咐」那些已经被繁重事务压得喘不过气的老师查看学生与 AI 的对话记录。
另外,多项研究表明,过度依赖聊天机器人会削弱批判性思维能力。Axios 在 2025 年报道了教师群体的普遍观察:越来越多的学生不愿意先自己思考,习惯直接向 AI 提问,然后把得到的答案照单全收。

▲ 图源:Aïda Amer/Axios
洛杉矶学区官员也明确表达了对 AI 摘要的顾虑:摘要浮现在搜索结果顶部,会让学生觉得难题已经应对,来源的多样性就此被压缩成一个声音。
感到困扰的也不止他们。不少人指出,搜索本想快速扫链接,却先被大块 AI 文字挡住,必须额外滚动或关闭,打断原本节奏。
「每次都要多点一下/多滑一下,累积下来特别累」。

Gartner 调查表明超过一半消费者对 AI 搜索不信任,61% 希望能关掉 AI 总结;实验研究也证实 AI Mode 降低了使用者满意度。

▲ 小红书搜索后自动弹出 AI 总结

▲ 微信搜一搜自动弹出 AI 回答
在 AI 之外,这件事也折射出更大的结构性难题。Google 在教育市面的位置,使它天然具备巨大的谈判长处。许多学区花了十年时间在 Google 生态平台里建立教学流程,这意味着即便他们不满意 Google 的某项决定,离开的成本也远高于留下来接受的代价。
Google 此次给予学校挑选权,但将这个挑选与搜索引擎的访问权捆绑在一起,使得本应清晰的教育决策变成了基础设施博弈,让那些希望谨慎的学区只能在能力缺失和原则妥协之间挑一个不情愿的答案。
本文来源:https://www.ifanr.com/1687912
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Gemini is set to become the AI that American students are compelled to use.
Recently, Google announced a policy adjustment that appears to be a concession: starting early next year, AI tools in its K-12 education accounts will be turned off by default, allowing school administrators to decide whether to enable capabilities like Gemini for students. Upon hearing the news, many educators breathed a sigh of relief, feeling they finally had some room for choice.
But the key detail was hidden in the "fine print," where the terms stated that "Search and AI are deeply integrated" and cannot be chosen separately.
In other words, once a school chooses to block AI, students will lose Google Search as well, with only a notice from the school displayed on the screen stating that the service is temporarily unavailable.
Education in this era certainly needs to embrace AI, but classrooms where students are forced to use a specific AI are already starting to change in character.
I feel like I'm being boxed in, Google.
Many school district administrators are caught in a dilemma. On one side are parents' and teachers' worries about the expansion of AI on campus; on the other is the unbearable weight of giving up a search engine.
Google's coverage in the U.S. K-12 education sector has long exceeded most people's everyday perception.
Google Workspace for Education covers over 170 million students and teachers, and Gmail, Docs, and Classroom form an inescapable operational platform for many school districts, handling everything from assigning homework to grading and communication. The search engine is even more of an infrastructure, existing by default like a whiteboard in a classroom.
Former Federal Trade Commission Chair Lina Khan believes that Google is "using its search monopoly to coerce schools" into accepting its AI models. Multiple courts have ruled that Google violates antitrust laws, and this move targeting schools precisely confirms that the company is "showing no restraint."
University of Southern California education professor Morgan Polikoff said this is a classic example of the tech industry "forcing AI onto everyone at all costs," even when the subjects have clearly refused.
From a business logic perspective, this bundling strategy is not hard to understand. The education market means not only tens of millions of users, but also the early shaping of habits. Letting a generation of students grow up in the Google ecosystem, using Gemini for homework and AI tools for previewing and reviewing, creates a much higher probability of converting them into paying users later in life than those who first encounter such models as adults.
The Defaults Have Changed
It is necessary to first review how it got to this point.
Before 2025, Google's AI capabilities in education were largely optional. Gemini was an add-on tool that had to be applied for separately, and school districts had clear switches to control it.
In August 2026, Google quietly did something: it enabled AI capabilities within Google Classroom by default for K-12 school districts that had previously granted students independent access to Gemini.
This move angered some school district administrators. They did not believe that their initial permission for students to access independent Gemini accounts was equivalent to authorizing the entire Classroom platform to enable AI tools by default.
Google subsequently announced a "reversal" of this decision, stating that it would restore school districts' autonomous control over AI in early 2027. But conditions came with it—turning off AI means losing Search as well. The reversal came with a price, and the price is borne by the school districts.
At the same time, Google announced that for school districts that allow AI to be enabled, there will be two search modes: one is an experience similar to the consumer version, including AI Overviews and AI conversations, and the other is a Gemini-based "Guided Learning" student version, which focuses on step-by-step guided learning.
Other search engines do not handle AI capabilities this way; DuckDuckGo provides a complete toggle for AI capabilities in its settings, allowing users to turn off Duck.ai and Search Assist, or to directly visit noai.duckduckgo.com for an AI-free search experience, which the company explicitly promotes as a difference from Google.
Brave Search also provides a setting to turn off "Answer with AI" with a single toggle. Although Microsoft Bing also has AI summary capabilities, it offers an official browser extension that allows users to toggle AI capabilities on or off with one click.
By contrast, Google Search's AI Overviews are officially classified as a core capability and cannot be turned off. The only workaround is to click the "Web" filter on the search results page or add the &udm=14 parameter when searching.
Who Will Protect Children's Minds
In September 2026, the two largest school districts in the United States—New York City and Los Angeles—successively announced broad restrictions on generative AI on school devices.
New York City has banned the use of AI for students in 8th grade and below. The Los Angeles Unified School District, however, hopes students can continue to use Google Search but with the AI summarization feature turned off, arguing that summaries will deter students from actively seeking out multiple information sources.
New York State later enacted a chatbot safety law that explicitly prohibits tech companies from providing various AI capabilities to minors, marking a more concrete step by a U.S. state in regulating AI in education.
Both school districts are clients of Google and, not coincidentally, are the most representative voices in this debate.
Backlash from parent groups is also steadily building; their concerns about AI are often an extension of the larger debate over screen time. AI has only pushed this debate to a new level of intensity.
▲ Source: Penn State Health News
The risks of children using AI can be understood on several levels.
The most direct risk is cheating; Gemini can handily complete a student's writing assignments. To counter this, Google launched the "Guided Learning" version for schools, claiming it will restrict the AI from directly providing answers, guide students through step-by-step thinking, and simulate the experience of one-on-one tutoring. However, tests have shown that even the consumer version of Guided Learning can generate a complete essay for an AP History question.
The school version allegedly has more restrictions, but in reality, it merely "instructs" teachers—who are already overwhelmed with heavy workloads—to review students' conversation logs with the AI.
Additionally, multiple studies indicate that over-reliance on chatbots can undermine critical thinking skills. Axios reported in 2025 on a common observation from teachers: a growing number of students are unwilling to think for themselves first and are accustomed to immediately asking AI for answers, which they then accept uncritically.
▲ Source: Aïda Amer/Axios
Los Angeles school district officials have also explicitly expressed their concerns about AI summaries: because the summaries appear at the top of search results, they make students feel that a difficult problem has already been solved, thereby reducing the diversity of sources to a single voice.
And they are not the only ones who are bothered. Many point out that they intend to quickly scan links in their search, but are first blocked by large blocks of AI-generated text, forcing them to perform extra scrolling or closing actions, which disrupts the original flow.
"You have to click or swipe one extra time every time, and the cumulative effect is exhausting."
A Gartner survey shows that over half of consumers distrust AI search, and 61% hope to be able to turn off AI summaries; experimental studies have also confirmed that AI Mode lowers user satisfaction.
▲ An AI summary automatically popping up after a search on Xiaohongshu (Little Red Book)
▲ An AI answer automatically popping up in WeChat's "Search"
Beyond AI, this issue also reflects a larger structural problem. Google's position in the education market inherently gives it immense bargaining power. Many school districts have spent a decade building their teaching workflows within the Google ecosystem, which means that even if they are dissatisfied with a particular Google decision, the cost of leaving is far higher than the price of staying and accepting it.
This time, Google offered schools a choice, but bundled this choice with access to its search engine, turning what should have been a clear educational decision into an infrastructure battle. This leaves cautious school districts to choose between an unwelcome answer of lacking capabilities or one of compromising their principles.