平台对 AI 生成的内容有什么要求

做出来的东西归谁，是工具的事。但下架帖子的是平台，而且它有自己的规则。差异比通常认为的更大：有的要求上传时加标记，不加就处罚；有的在规则中根本不提 AI；还有的干脆禁止自动生成的内容。

下面是各平台自身规则中所写的内容。这个主题的规则比其他主题变化更频繁。

Facebook

是否需要标记。 对照片级真实感的视频和逼真的音频有此要求。原文：«We require people to disclose, using our AI-disclosure tool, whenever they post organic content with photorealistic video or realistic-sounding audio that was digitally created or altered, and we may apply penalties if they fail to do so.»对图片则没有这样的要求：平台承诺自己给它们加标记该规则对 Facebook、Instagram 和 Threads 通用——Meta 将其归在《社区准则》的 «Misinformation» 一节。平台自己的帮助中心文章（«Label AI content on Instagram»，help.instagram.com/761121959519495）无法用普通请求访问：不带浏览器时域名返回 HTTP 400，且 help.instagram.com 的 robots.txt 含有 «User-agent: ClaudeBot / Disallow: /» 来源, 已核查 2026-08-04

用什么方式标记。 标记叫做«AI info»（此前是«Made with AI»）：«Meta began adding labels to a wider range of video, audio and image content»（自2024年5月起），«These are displayed as “AI Info”»。创作者使用的是单独的工具：«we’re adding a feature for people to disclose when they share AI-generated video or audio so we can add a label to it»。平台会显示标记的来由：«shares whether the content is labeled because of industry-shared signals or because someone self-disclosed» 关于披露工具的表述出自 Meta 博客 https://about.fb.com/news/2024/02/labeling-ai-generated-images-on-facebook-instagram-and-threads/。可访问来源中没有逐步操作指南（发布表单里的开关具体在哪）：Meta 帮助页面无法访问，见 ai_disclosure_required 的注释来源, 已核查 2026-08-04

平台是否自行识别。 是，通过行业元数据实现。原文：«we’re building industry-leading tools that can identify invisible markers at scale – specifically, the “AI generated” information in the C2PA and IPTC technical standards»；«in the coming months we will label images that users post to Facebook, Instagram and Threads when we can detect industry standard indicators that they are AI-generated»。另有：«Our “Made with AI” labels on AI-generated video, audio and images will be based on our detection of industry-shared signals of AI images or people self-disclosing that they’re uploading AI-generated content» 两个标准 C2PA 和 IPTC 均被直接点名。第二条引文出自 https://about.fb.com/news/2024/04/metas-approach-to-labeling-ai-generated-content-and-manipulated-media/。Meta 承认统计并不完整：«labeling methodology is still evolving»，且数据 «may not capture labels on some content that is edited with AI» 来源, 已核查 2026-08-04

禁止什么。 没有对 AI 内容的直接禁令——只有加标和对危害的禁止。原文：«We may place an informative label on the face of content – or reject content submitted as an advertisement – when the content is a photorealistic image or video, or realistic sounding audio, that was digitally created or altered and creates a particularly high risk of materially deceiving the public on a matter of public importance.» 无论制作方式如何都会被删除：«We remove misinformation or unverifiable rumors that expert partners have determined are likely to directly contribute to a risk of imminent violence or physical harm to people»；«We consult with leading health organizations to identify health misinformation likely to directly contribute to imminent harm to public health and safety»；关于选举和人口普查——«We remove misinformation that is likely to directly contribute to a risk of interference with people’s ability to participate in those processes» Meta 并没有专门针对 AI 的“不得伪造真人声音／面孔／选举／医疗声明”式清单：禁令按危害表述，而非按制作方式。因此判定为 partial 来源, 已核查 2026-08-04

违规会有什么后果。 «…and we may apply penalties if they fail to do so»——这是关于未加标处罚的唯一表述。具体如何处罚（下架、限流、取消变现资格），规则并未说明 规则中既没有期限，也没有分级和门槛——只有«may apply penalties»。在 Meta 可访问的资源上没有找到更详细的、关于未加标注后果的页面 来源, 已核查 2026-08-04

能否借此获得收益。 未直接提及 AI；规则针对的是非原创内容：«Content that is duplicative or involves minor edits to another creator’s post» 被视为 «unoriginal and deprioritized»；«Creators who are not adding substantial creative value will see their content deprioritized, and they may see their account deemed non-recommendable and demonetized, if they continue to post primarily unoriginal content» Meta 这篇关于内容原创性的博客文章中没有出现 «AI» 一词——即平台并不因使用 AI 本身而处罚，而是处罚非原创。«Content Monetization Policies» 本身无法核验：它位于 www.facebook.com/business/help，该域名被 robots.txt 整体封禁（«User-agent: * / Disallow: /»）来源, 已核查 2026-08-04

平台卡片

Instagram

平台卡片

用什么方式标记。 标记由平台加上，而不是创作者：«LinkedIn members will start seeing a new icon that helps members trace the origin of AI-created media, including the source and history of the content, and whether it was created or edited by AI» 这是 LinkedIn 新闻中心2024年5月15日的官方消息（«LinkedIn adopts C2PA standard»）。在可访问的页面上没有描述创作者上传时的勾选框。帮助文章«Content credentials»位于 www.linkedin.com/help/linkedin/answer/a6282984——该域名是关闭的，参见 ai_disclosure_required 的注记来源, 已核查 2026-08-04

平台是否自行识别。 是，依据 C2PA 标准：«Today we shared in a post from Patrick Corrigan, our VP of Legal, how we are rolling out the new C2PA AI-generated content standard on the LinkedIn platform»；该图标显示«the source and history of the content, and whether it was created or edited by AI» 与主域名不同，新闻中心 news.linkedin.com 是开放的（robots.txt 为空）。可访问的页面上没有提到 IPTC，也没有提到自有的识别器 来源, 已核查 2026-08-04

禁止什么。 禁令是围绕危害而不是围绕制作方式表述的。按 LinkedIn 官方规则页面的搜索索引，平台会删除«Synthetic or manipulated media, such as doctored images or videos that distort real-life events and likely to cause harm to the subject, other individuals or groups, or society at large»；通用规则是«Do not share content that is false, misleading, or intended to deceive» www.linkedin.com 域名在 robots.txt 中对我们关闭：那里以单独的分组列出了«User-agent: anthropic-ai»、«ClaudeBot»、«Claude-Web»、«Claude-User»、«Claude-SearchBot»，每一组都带«Disallow: /»。LinkedIn 的社群守则和帮助中心只在这个域名上，不能读，也不能绕过禁令。所引文字取自这同一官方页面的搜索索引；没有在页面本身上复核逐字准确性——因此为 partial 来源, 已核查 2026-08-04

平台卡片

是否需要标记。 创作者没有标注 AI 内容的义务——标记由平台自己加上。创作者的标示被称为可以做，而不是必须做：«Some Pins may also have the AI modified label if the content owner indicated that the Pin was generated or modified using AI»。在《AI 可接受使用规则》中，Pinterest 完全没有关于标注的要求已核查三个官方页面：关于标记的帮助文章（help.pinterest.com/en/article/gen-ai-labels）、社群守则（policy.pinterest.com/en/community-guidelines）和 AI 使用规则（policy.pinterest.com/en/genai-acceptable-use-guidelines）。没有任何一处写着“必须标注”。通过 WebFetch 读取：两个域名的 robots.txt 中禁止了«anthropic-ai»、«ClaudeBot»、«Claude-Web»，但没有禁止«Claude-User» 来源, 已核查 2026-08-04

用什么方式标记。 标记叫做«AI modified»，由平台自己加上：用户«in close-up at the bottom left after you click on the Pin»看到它。对广告 Pin，披露在省略号图标下面，表述为«was modified with AI»或«may be modified with AI»。创作者可以对标记提出异议：要申辩，需要«contact our support team» 帮助中心既没有描述上传时的勾选框，也没有描述创建 Pin 表单里的“这是 AI 做的”字段 来源, 已核查 2026-08-04

平台是否自行识别。 是，有两种方式：«Pinterest analyzes a Pin’s metadata following the IPTC Metadata Standard, but also developed classifiers that automatically detect Gen AI content, even if the content doesn’t have obvious markers» 点名的正是 IPTC 标准（不是 C2PA）。此外还有不依赖元数据的自有识别器。平台的新闻页面（newsroom.pinterest.com/news/introducing-gen-ai-labels/）上说法相同 来源, 已核查 2026-08-04

禁止什么。 通用规定：«Our policies apply to everyone and to all types of content, including synthetically generated content»。禁止的有：«Content that is synthetically generated or manipulated to degrade and/or shame»（霸凌）；«synthetically generated or manipulated content that impersonates a person or organization»——对此«We’ll deactivate accounts»；关于选举：«Fabricated or manipulated content (e.g. content generated with AI tools) making someone appear to do or say something they didn’t with the clear intent of influencing public sentiment about a political figure and/or election»；«Instructions, tools or services for creating sexually explicit or nudity content (e.g. “nudifiers”)» Pinterest 对选举的禁令表述得很直接，而且直接点到了 AI 工具。没有针对 AI 作出的医疗声明的单独禁令——只有不与制作方式挂钩的、对医疗虚假信息的通用禁令来源, 已核查 2026-08-04

平台卡片

禁止什么。 没有对 AI 内容的单独禁令。最接近的是规则 5：«Be authentic. You don’t have to use your real name, but do not intentionally mislead others or impersonate an individual or entity in a deceptive manner»；以及关于«content manipulation»的规则 2 关于冒充他人的规则详情页（support.reddithelp.com/hc/articles/41180423371156）被 Cloudflare 关闭（HTTP 403），因此它是否提及深度伪造未能核查。Reddit 规则已在 redditinc.com/policies/reddit-rules 通读（该域的 robots.txt 对所有人放行 /policies 栏目）。规则的详细解释位于 support.reddithelp.com——那里是 Cloudflare 防护、HTTP 403 和 «Just a moment... Enable JavaScript and cookies to continue»；www.reddit.com 被 «User-agent: * / Disallow: /» 规则关闭。不得绕过。另注：redditinc.com 的自动阅读被关闭，页面以普通 GET 获取。来源, 已核查 2026-08-04

平台卡片

Snapchat

是否需要标记。 对普通发布没有这项义务；对可变现的发布则有。变现规则原文：«Content that leverages sophisticated editing or AI-based tools is monetizable—if it is original, entertaining, or informative, not misleading, and if the use of AI is disclosed anywhere in the content or creator profile.»《社群守则》中没有标注 AI 的要求，只有一句笼统的：«We also expect Snapchatters to use AI responsibly» 披露可以放在«anywhere in the content or creator profile»——也就是发布内容本身或账号简介中的任何位置，不需要单独的字段。社群守则已按 https://values.snap.com/privacy/transparency/community-guidelines 地址核查来源, 已核查 2026-08-04

用什么方式标记。 规则中没有描述上传时的单独勾选框：披露是用自己的话作出的——«disclosed anywhere in the content or creator profile»。对用 Snapchat 自家工具做出的 AI 内容，标记由平台加上：«AI-generated content that was created within Snapchat is eligible for recommendation. Our tools always include transparency indicators» 用 Snapchat 自家工具做出的内容上，«transparency indicators»（透明度标识）究竟是什么样子，已核查的页面上没有描述 来源, 已核查 2026-08-04

禁止什么。 社群守则：«We prohibit spreading false information that causes harm or is malicious, such as denying the existence of tragic events, unsubstantiated medical claims, undermining the integrity of civic processes, or manipulating content for false or misleading purposes (whether through generative AI or through deceptive editing)»；关于霸凌——«sharing humiliating imagery (including manipulated or AI-generated imagery)»；通用条款——«We reserve the right to take appropriate enforcement action against accounts that use AI to violate our Community Guidelines or abuse our AI products. This can include termination of an account»。变现规则另外禁止«Deceptively manipulated media, including content that has been generated or enhanced with the use of AI tools to manipulate users (for example, editing an image of someone’s body or face to imply a radical transformation…)» 医疗声明（«unsubstantiated medical claims»）和破坏«integrity of civic processes»（即选举）与 AI 制作被直接并列写在一起。最后一段引文出自 https://values.snap.com/policy/creator-monetization-policy 来源, 已核查 2026-08-04

违规会有什么后果。 对借助 AI 违反守则的行为：«We reserve the right to take appropriate enforcement action against accounts that use AI to violate our Community Guidelines or abuse our AI products. This can include termination of an account.»没有专门针对缺少标注的处罚；后果是失去变现资格，因为披露 AI 是可变现的条件之一 Snap 的页面上没有“没标注就下架／限流”这样的直接对应；有的是“没披露就不能变现” 来源, 已核查 2026-08-04

能否借此获得收益。 要赚钱必须同时满足三个条件：«Content that leverages sophisticated editing or AI-based tools is monetizable—if it is original, entertaining, or informative, not misleading, and if the use of AI is disclosed anywhere in the content or creator profile.»与此同时，完全在 Snapchat 之外用 AI 做出的内容展示会更差：«Our content ranking algorithm rewards authentic, human-made content over wholly AI-generated content created outside of Snapchat, even when AI-generated content has transparency disclosures. AI-generated content that was created within Snapchat is eligible for recommendation.»不予变现的是«content where the core visual or audio contribution is mechanically or automatically generated» 一个重要细节：标注并不能免于在推荐中被降权——«even when AI-generated content has transparency disclosures»。关于推荐的规则见 https://values.snap.com/policy/content-guidelines-recommendation-eligibility/recommendation-eligibility/quality 。变现取决于是否适合被推荐：«If you primarily or frequently publish content that is “not eligible for recommendation,” you are most likely not a good candidate for content monetization on Snapchat» 来源, 已核查 2026-08-04

平台卡片

Threads

是否需要标记。 适用 Meta 的通用规则：«We require people to disclose, using our AI-disclosure tool, whenever they post organic content with photorealistic video or realistic-sounding audio that was digitally created or altered, and we may apply penalties if they fail to do so.»Threads 被直接列入会加标记的应用之中：«in the coming months we will label images that users post to Facebook, Instagram and Threads when we can detect industry standard indicators that they are AI-generated» 没有找到 Threads 自己关于 AI 的规则：Threads 没有自己的“社群守则”，适用的是 Meta 的守则。它自己的帮助文章«Label AI content on Threads»（help.instagram.com/407718162047721）无法访问：不用浏览器时该域名返回 HTTP 400，help.instagram.com 的 robots.txt 中有«User-agent: ClaudeBot / Disallow: /»。因此为 partial：规则在 Meta 层面得到确认，但没有在 Threads 自己的页面上得到确认来源, 已核查 2026-08-04

用什么方式标记。 有«AI info»标记，创作者也有单独的披露工具：«we’re adding a feature for people to disclose when they share AI-generated video or audio so we can add a label to it»；«Meta began adding labels to a wider range of video, audio and image content»，«These are displayed as “AI Info”» 透明度中心的页面把«Facebook, Instagram, and Threads»放在一起说。Threads 发布表单中的开关究竟是什么样子，没有核查过，Threads 的帮助中心是关闭的（参见 ai_disclosure_required 的注记）来源, 已核查 2026-08-04

平台是否自行识别。 是。«we’re building industry-leading tools that can identify invisible markers at scale – specifically, the “AI generated” information in the C2PA and IPTC technical standards»；«in the coming months we will label images that users post to Facebook, Instagram and Threads when we can detect industry standard indicators that they are AI-generated» Threads 出现在 Meta 博客文章的标题本身之中。两项标准都被提及——C2PA 和 IPTC 来源, 已核查 2026-08-04

禁止什么。 没有针对 AI 的单独禁令清单；适用的是 Meta 的守则：«We may place an informative label on the face of content – or reject content submitted as an advertisement – when the content is a photorealistic image or video, or realistic sounding audio, that was digitally created or altered and creates a particularly high risk of materially deceiving the public on a matter of public importance»；无论制作方式如何，带有暴力风险的虚假信息、有害的医疗虚假信息，以及关于选举和人口普查的虚假信息都会被删除 Threads 没有自己关于 AI 的禁令清单；上面所说的全是 Meta 的规则，适用于该应用 来源, 已核查 2026-08-04

违规会有什么后果。 «…and we may apply penalties if they fail to do so»——这是 Meta 关于未加标注将受罚的通用表述；究竟如何处罚没有披露 没有找到专门针对 Threads 描述的处罚 来源, 已核查 2026-08-04

平台卡片

TikTok

是否需要标记。 对写实类 AI 内容有此要求。平台博客原文：«The policy requires people to label AI-generated content that contains realistic images, audio or video, in order to help viewers contextualize the video and prevent the potential spread of misleading content»；«we label AIGC made with TikTok AI effects, and have required creators to label realistic AIGC for over a year» 该要求的原始出处是社群守则中的«Integrity and authenticity»一节，地址为 https://www.tiktok.com/community-guidelines/en/integrity-authenticity 。这一页读不到：www.tiktok.com 域名的 robots.txt 中有«User-agent: ClaudeBot»、«Claude-User»、«Claude-SearchBot»、«anthropic-ai»几组，都带«Disallow: /»。帮助文章 support.tiktok.com/en/using-tiktok/creating-videos/ai-generated-content 以 302 跳转到同一个被关闭的域名。因此规则的逐字表述取自平台自己的博客来源, 已核查 2026-08-04

用什么方式标记。 发布时有单独的标注工具，同时也允许用手边的办法：«we're rolling out a new tool for creators to easily inform their community when they post AI-generated content. The new label… they can apply it to any content that has been completely generated or significantly edited by AI»；«Creators can now do this through the new label (or other types of disclosures, like a sticker or caption)»。关于覆盖面：«We also built a first-of-its-kind tool to make this easy to do, which over 37 million creators have used since last fall» 平台的公开来源中没有分步说明（“该点哪里”）：帮助中心跳转到 www.tiktok.com，而该域名在 robots.txt 中对 Claude 的代理是禁止的。«sticker or caption»——也就是贴纸或字幕——被直接写成勾选框之外的可行替代来源, 已核查 2026-08-04

平台是否自行识别。 是，有三种方式。C2PA：«we’re expanding auto-labeling to AIGC created on some other platforms by launching the ability to read Content Credentials, a technology from the Coalition for Content Provenance and Authenticity (C2PA)… TikTok is the first video sharing platform to put Content Credentials into practice»。自有标记：«we’ll also start attaching Content Credentials to TikTok content, which will remain on content when downloaded»。隐形水印：«we’ll start adding invisible watermarks to AI-generated content made with TikTok tools like AI Editor Pro, and content uploaded with C2PA Content Credentials»；«“Invisible watermarks” add another layer of safeguards with a robust technological “watermark” that only we can read, making it harder for others to remove»。标记会显示来源：«we enhanced our AIGC labels by adding more context around whether content was labeled due to our AI detection, creator labels, or TikTok AI tools» 关于隐形水印和标记来源的内容在平台博客的第二个页面上：https://newsroom.tiktok.com/en-us/more-ways-to-spot-shape-and-understand-ai-content 。平台在那里还给出了一个数字：超过30亿条视频被标记为 AI 内容来源, 已核查 2026-08-04

禁止什么。 «our policies firmly prohibit harmfully misleading AI-generated content—whether it's labeled or not»；«we… signed an industry pact this year to combat the use of deceptive AI in elections»。还有：«We continue to refine our labeling approach as industry norms and expert guidance evolve, while maintaining firm policies against harmful AIGC» 未能以公开形式取得禁令清单（真人声音、面孔、选举、医疗声明）：它在 www.tiktok.com 上的社群守则里，而该站在 robots.txt 中对 Claude 的代理是关闭的（«Disallow: /»）。«whether it’s labeled or not»这一表述很重要：标注并不解除禁令来源, 已核查 2026-08-04

平台卡片

Twitch

禁止什么。 禁止滥用伪造内容：«Creating “deepfake” videos to insult or abuse another person»（在关于霸凌的一节中，紧邻«Sharing negative doctored or artistic content to abuse or degrade another person»）；«Creating or sharing doctored or ‘deepfake’ intimate images of an identifiable person»（在关于性骚扰的一节中）。关于真人声音、选举和医疗声明与 AI 的关联则只字未提；有的是不与制作方式挂钩的、对«Impersonation of another person, company, or organization»和«Harmful Misinformation Actors»的通用禁令 «deepfake»这个词是整份社群守则中对 AI 的唯一提及。两条禁令都是围绕对特定个人的伤害来表述的，而不是围绕制作方式 来源, 已核查 2026-08-04

平台卡片

X (Twitter)

是否需要标记。 没有发现要求给所有 AI 内容加标注的通用规定。只有一种情形有此要求——武装冲突：«For any AI generated content depicting armed conflict, you must add a clear disclosure that it is AI generated» 该页面被 Cloudflare 防护挡住：curl 和 WebFetch 都得到 HTTP 403 和占位页«Just a moment... Enable JavaScript and cookies to continue»。不能绕过这类防护。所引文字取自针对这同一官方页面的搜索索引，没有在页面本身上复核逐字准确性——因此 confidence 为 partial。该要求属于“内容变现标准”，也就是针对赚钱的人，而不是针对所有创作者来源, 已核查 2026-08-04

用什么方式标记。 没有描述创作者在发布时可用的单独标注工具：规则要求«add a clear disclosure that it is AI generated»，也就是在帖子本身之内作出说明。标记由平台加上：«For posts containing media determined to have been significantly and deceptively altered or fabricated, X may add a “Manipulated media” label» 该页面被 Cloudflare 防护挡住：curl 和 WebFetch 都得到 HTTP 403 和占位页«Just a moment... Enable JavaScript and cookies to continue»。不能绕过这类防护。所引文字取自针对这同一官方页面的搜索索引，没有在页面本身上复核逐字准确性——因此 confidence 为 partial。在可访问的 X 页面上，既没有描述上传时的勾选框，也没有描述元数据字段来源, 已核查 2026-08-04

平台是否自行识别。 部分是：«To determine if media have been significantly and deceptively altered or fabricated, X may use its own technology or receive reports through partnerships with third parties»。没有找到关于 C2PA 或 IPTC 的提及该页面被 Cloudflare 防护挡住：curl 和 WebFetch 都得到 HTTP 403 和占位页«Just a moment... Enable JavaScript and cookies to continue»。不能绕过这类防护。所引文字取自针对这同一官方页面的搜索索引，没有在页面本身上复核逐字准确性——因此 confidence 为 partial。在可访问的 X 页面上，关于内容来源标准（C2PA、IPTC）什么也没有说来源, 已核查 2026-08-04

禁止什么。 «Media that is shared as authentic but is significantly and deceptively altered, manipulated, or fabricated in a way that fundamentally changes its meaning is prohibited, including media depicting a real person that has been fabricated or simulated, especially through use of algorithms or broader artificial intelligence»。不予处罚的情形：«X will not take action to label or remove media that have been edited in ways that do not fundamentally alter their meaning, such as retouched photos or color-corrected videos» 该页面被 Cloudflare 防护挡住：curl 和 WebFetch 都得到 HTTP 403 和占位页«Just a moment... Enable JavaScript and cookies to continue»。不能绕过这类防护。所引文字取自针对这同一官方页面的搜索索引，没有在页面本身上复核逐字准确性——因此 confidence 为 partial。禁令是围绕欺骗而不是围绕制作方式来表述的：关键在于帖子被当作真实内容发出（«shared as authentic»）来源, 已核查 2026-08-04

违规会有什么后果。 对武装冲突题材的 AI 视频未加标注：«Effective March 3, 2026 users who post AI-generated videos of an armed conflict without adding a disclosure that it was made with AI will be suspended from Creator Revenue Sharing for 90 days. Subsequent violations will result in a permanent suspension from the program»。对具有误导性的篡改内容，平台«may add a “Manipulated media” label»，也会给帖子加标记或将其下架该页面被 Cloudflare 防护挡住：curl 和 WebFetch 都得到 HTTP 403 和占位页«Just a moment... Enable JavaScript and cookies to continue»。不能绕过这类防护。所引文字取自针对这同一官方页面的搜索索引，没有在页面本身上复核逐字准确性——因此 confidence 为 partial。这是在该平台上找到的唯一一项针对缺少 AI 标注的数值化处罚：90天内被剥夺收入分成，重犯则永久剥夺来源, 已核查 2026-08-04

能否借此获得收益。 «Monetized content, like other content on X, must be in compliance with X’s rules on Safety, Authenticity, and Privacy. You must only monetize content that is original and authentic, or otherwise content that you have the rights to monetize»。此外，对涉及武装冲突的 AI 内容有强制披露要求（参见 ai_disclosure_required）该页面被 Cloudflare 防护挡住：curl 和 WebFetch 都得到 HTTP 403 和占位页«Just a moment... Enable JavaScript and cookies to continue»。不能绕过这类防护。所引文字取自针对这同一官方页面的搜索索引，没有在页面本身上复核逐字准确性——因此 confidence 为 partial。没有发现禁止靠 AI 内容赚钱的直接规定——要求的是原创性、真实性，以及在指定情形下的披露来源, 已核查 2026-08-04

平台卡片

YouTube

是否需要标记。 对照片级真实感的 AI 内容有此要求。原文：«we require creators to disclose when they use AI to meaningfully alter or generate photorealistic content»；«Creators must disclose GenAI content that: Makes a real person appear to say or do something they didn’t do. Alters footage of a real event or place. Generates a realistic scene that didn’t actually occur.»不需要标注的情形：«Creators don’t need to disclose non-realistic content that’s made with AI, or edits to realistic content that are minor. Minor edits are ones that are primarily aesthetic, and don't alter the content in a way that could mislead the viewer about what actually happened.»无需标注的例子：«Applying beauty filters»、«Color adjustment or lighting filters»、«Production assistance, like using generative AI tools to create or improve a video outline, script, thumbnail, title, or infographic»、«Caption creation»、«Cloning one’s own voice to create voice overs or dubs»。需要标注的例子：«AI generated music»、«AI generated realistic videos of a match between two real professional tennis players»、«Making it appear as if someone gave advice that they did not actually give»、«Depicting a public figure stealing something they did not steal» 页面已改名：标题现在是«Disclosing use of GenAI content»（此前是«altered or synthetic content»）。页面上的示例清单被标注为不完整（«this isn’t a complete list»）。通过普通 GET 获取，页面是开放的来源, 已核查 2026-08-04

用什么方式标记。 在 YouTube Studio 上传时的勾选项：«In the Attributes section, under “AI use,” select Yes if your content meets the disclosure requirements or No if not.»观众看到的标记形式：«Label on the video player for photorealistic AI content»；«Labels may appear in the expanded description for AI content that is non-photorealistic or animated» 标记由创作者本人在电脑端或移动应用的 Studio 中的«Attributes»（属性）一步设置。除«AI disclosure survey»问卷之外，没有关于取消标记的单独帮助文档 来源, 已核查 2026-08-04

平台是否自行识别。 是。«YouTube may automatically apply an AI label on the video player or in the expanded description for: Content made using YouTube's GenAI tools; Content that contains C2PA metadata; Content that our internal systems detect is Al generated or altered.»自动标记并不总能取消：«However, content made with YouTube’s AI tools, content containing C2PA metadata, or content labeled after manual review can not be adjusted.» 明确点名了 C2PA 元数据和自有的识别系统。关于自动标记出错：«If our systems make an error, creators will have the ability to change AI disclosure in most cases by selecting No in the AI disclosure survey» 来源, 已核查 2026-08-04

禁止什么。 标记并不解除禁令：«Creators should disclose when content is generated or meaningfully altered with AI, but this disclosure is not a free pass to impersonate a person, entity, or channel. Our Community Guidelines, including our impersonation policy, apply to all content regardless of how it is made.»禁止的行为：«It may also include using AI to copy the voice or likeness of an individual, to make it appear as if the channel is owned or authorized by that individual»；«Using someone's AI-generated likeness or voice to falsely imply they own or authorize your channel’s content»；«Using AI to make it look like a famous person is willfully participating in your video or endorsing a product»；«Misrepresenting an individual by using an AI version of their voice and false metadata to pass off a fake speech as real» 这是关于冒充他人的规则（impersonation policy）。受害者可以提出申诉：«If you find your personal information, a video featuring you, or AI-generated content mimicking your likeness or voice without your consent, you can file a request for removal under our privacy guidelines»。该页面上没有针对选举类 AI 内容的单独禁令——它在关于选举虚假信息的规则里来源, 已核查 2026-08-04

违规会有什么后果。 «Creators who consistently choose not to disclose this information may be subject to manual application of a label, or penalties from YouTube, including removal of content or suspension from the YouTube Partner Program.»而标注本身不会受到处罚：«Disclosing AI content won’t limit a video’s audience or impact its eligibility to earn money.» 所描述的处罚针对的是系统性地（«consistently»）不标注，而不是单次情况。页面没有给出具体的期限和门槛 来源, 已核查 2026-08-04

能否借此获得收益。 如果 AI 内容是原创的、不是套模板的，就可以从中获利。允许：«Content that expresses your unique creative voice, like using AI to visualize a unique character and narrative you invented»；«Content that utilizes creative tools to assist in delivering a unique, well-researched, or creative narrative, like using AI to edit your video scripts or generate a unique background visual for your content»。不允许变现：«AI-generated content made with generic or unoriginal templates giving the impression of mass production without adding the creator’s original, authentic insights or perspective»；«Content that lacks a clear narrative arc or logical progression, such as videos that stitch together unrelated or inconsistent AI clips to surprise or shock viewers»；«Content that uses deceptive or misleading imagery or narratives, such as realistic visuals tricking viewers into believing a fake celebrity death or natural disaster has occurred»。另有一条单独规则«AI Personas Related to Sensitive Topics»：«This policy refers to channels that use AI-generated personas to deliver information on sensitive topics… on topics such as health, legal issues, finances, or politics… channels uploading this content will not be allowed to monetize»——例子：«An AI “doctor” providing medical diagnoses, health advice, or wellness remedies»、«AI-generated podcast hosts offering financial guidance, investment tips, or wealth management advice»、«AI personas giving legal advice or interpreting laws» 同一页面的变更日志中：«July 15, 2025: We’re making a minor update to our “repetitious content” policy to better clarify this includes content that is repetitive or mass-produced. We are also renaming this policy from “repetitious content” to “inauthentic content.”»关于 AI 标注不影响收入的那句话，见 support.google.com/youtube/answer/14328491 来源, 已核查 2026-08-04

平台卡片

Dzen

禁止什么。 被禁止的是自动生成的内容本身，与是否标注无关：Dzen 禁止“完全或几乎完全由自动方式生成、没有作者参与”的内容。清单中包括“未经编辑的其他语言自动翻译”“自动生成的文本”“自动生成的视频配音和音轨”，以及含有“彼此无关的重复词句”的材料。这类内容被归入垃圾信息——属于“严重违规” 这是遇到过的最严格的一种处理方式：不是“请标注”，而是“不可以”。Dzen 的帮助中心是在浏览器中生成的：普通 GET 只返回约 1 KB 的空壳，没有正文，因此这些页面是通过 WebFetch 读取的。dzen.ru 的 robots.txt 没有禁止这些地址。来源, 已核查 2026-08-04

违规会有什么后果。 对自动生成的内容：这类发布“只向订阅者展示”；再进一步——“Dzen 可以封禁该发布内容甚至整个频道——它们对任何人都不可见，即使通过直接链接也看不到”；“垃圾信息大量存在的频道和网站在平台上不可访问” 处罚的对象不是缺少标注（本来也不要求标注），而是自动生成这一事实本身。Dzen 的帮助中心是在浏览器中生成的：普通 GET 只返回约 1 KB 的空壳，没有正文，因此这些页面是通过 WebFetch 读取的。dzen.ru 的 robots.txt 没有禁止这些地址。来源, 已核查 2026-08-04

能否借此获得收益。 变现条件中没有关于 AI 的单独规则。通用规定：“如果发布内容违反要求，Dzen 可以关闭其变现。如果这类违规很多——限制将扩展到整个频道”。由于自动生成的内容被归入垃圾信息并被禁止，靠它是赚不到钱的变现规则页面已阅读：上面没有关于神经网络、AI 或自动生成内容的字眼。无法赚钱这一结论来自垃圾信息规则中的禁令，而不是平台自己的直接表述——因此标为 partial 来源, 已核查 2026-08-04

平台卡片

该怎么做

拿不准就加标记。加标记在任何地方都不会受罚，不加标记在很多地方都会受罚。
没有规则不等于获得许可。 规则可能明天才出现，并对昨天发布的内容生效。
平台标记与广告披露是两回事。前者说的是“这是机器做的”，后者说的是“有人为此付了钱”。对广告披露的要求是单独的。
请核对变现规则，而不只是发布规则。 部分平台允许发布 AI 内容，但不允许用它变现。

另请参阅：用 AI 创作内容 · 收入 · 平台限制