What each platform asks you to label

Why platform rules come first

For most small businesses, the first AI disclosure rule they actually run into is not a statute. It is a checkbox in YouTube Studio, a product-feed attribute in Google Merchant Center, or a question in the Kindle publishing flow. Platforms enforce faster than regulators: an ad gets rejected the same day, a listing disappears, a channel loses monetization. The good news is that their rules are mostly narrow and written down. Here is what each one asks for, with a link to each platform's own policy page. Where we could not confirm a detail on that page, we say so.

Quick comparison

Google Search

Google does not require you to label AI-written web pages. Its guidance on generative AI content (updated October 1, 2026) says AI content is fine if it is helpful and accurate, and only suggests that publishers "consider adding information on how your content was created" where readers would expect it. It also recommends image metadata for AI images, and tells publishers to fact-check all AI output, including page titles, meta descriptions, structured data and image alt text.

What Google does prohibit is scaled content abuse: using AI (or anything else) to churn out many pages that add no value for readers. That spam policy took effect March 5, 2024. Sites that break it can rank lower or drop out of results, including through manual actions.

For a small business: an AI-assisted service page that a person checked and that answers real customer questions is fine. Fifty near-identical "plumber in [town]" pages generated in an afternoon is the pattern that gets sites in trouble. A short "how we made this" note is optional but costs nothing; see the blog post example.

Google Merchant Center and Shopping

This is the one Google product with a hard labeling rule. If you upload products to Merchant Center for Shopping ads or free listings:

• AI-generated product images must carry IPTC DigitalSourceType metadata set to trainedAlgorithmicMedia. Google also asks you to keep the compositeSynthetic and algorithmicMedia values where they exist. This applies to your main image, additional images and lifestyle images. Do not strip embedded metadata, including from images made with Google's own Product Studio.
• AI-generated titles and descriptions go in the structured_title and structured_description attributes instead of the regular title and description, each with digital_source_type set to trained_algorithmic_media and the text in content.

In plain terms: many AI image tools already write this tag into the image file, so your main job is not to strip it. If your tool does not add it, add it with a metadata editor before you upload. Do not re-save AI images through editors, compressors or screenshot tools that drop file data. To check a file, open it in any online metadata viewer and look for "Digital Source Type". For AI-written titles and descriptions, ask whatever sends your products to Google (your Shopify or WooCommerce Google app, or your feed tool) how to fill the "structured title" and "structured description" fields.

Google has not published an effective date or a specific penalty for this rule. General Merchant Center enforcement may apply. Google's Search guidance repeats that this labeling is a "must" for ecommerce product data.

Google Ads (including YouTube ads)

Ordinary product ads

There is no general AI-disclosure requirement for regular commercial ads. But the Misrepresentation policy (updated March 2024) bans ads that try to get money or information by impersonating, or falsely implying affiliation with or endorsement by, a public figure, brand or organization. A disclaimer does not cure it. Google treats this as egregious: the account is suspended on detection, without warning, and you cannot advertise with Google again. This is the rule that catches AI deepfake "celebrity endorsement" ads.

Election ads

Verified election advertisers must check the "Altered or synthetic content" box in campaign settings when an ad has synthetic or digitally altered images, video or audio that show real or realistic-looking people or events in a way that is not authentic. Google auto-generates the in-ad disclosure on some formats (mobile Feeds and Shorts, and in-stream video on all screens). On other formats you add your own clear disclosure, such as "This image does not depict real events." Inconsequential edits like resizing, cropping, color correction and red-eye removal are exempt. Ads that break the rule are disapproved, and Google warns at least seven days before any account suspension. Political content policy.

No political ads in the EU

Since October 2025, Google does not serve political advertising in the EU at all, in response to the EU's political-ads regulation. The definition is broad and can catch issue or cause ads. Advertisers must self-declare whether campaigns contain EU political ads, and since April 1, 2026 accounts that have not declared can be blocked from editing campaigns through the API. Google's announcement.

YouTube

YouTube requires creators, business channels included, to disclose realistic altered or synthetic content. You do it at upload by answering "Yes" under "AI use" in the video's attributes. Disclosure is required when the content seems realistic and, for example:

• makes a real person appear to say or do something they did not;
• alters footage of a real event or place;
• shows a realistic scene that never happened; or
• uses AI music as the main focus of the video.

You do not need to disclose clearly unrealistic or animated content; minor edits like beauty filters, color, lighting or blur; AI help with scripts, outlines, titles, thumbnails or ideas; captions, upscaling and audio repair; cloning your own voice for voiceovers or dubbing; or gameplay footage.

Since May 2026, YouTube also detects significant photorealistic AI use itself and labels undeclared content. On long videos the label now sits below the player; on Shorts it is an overlay. You cannot remove a label from content made with YouTube's own AI tools, content carrying Content Credentials (C2PA) metadata showing it is fully AI-generated, or content labeled after a manual review.

Disclosing does not hurt reach or monetization. Consistently failing to disclose can lead to a label being added, removal, or suspension from the YouTube Partner Program. Separately, YouTube reportedly stopped monetizing mass-produced, repetitive content from July 15, 2025.

Facebook, Instagram and Threads

Organic posts

Meta requires you to use its AI disclosure tool when you post photorealistic video or realistic-sounding audio that was digitally created or altered. Meta also applies an "AI info" label on its own when it detects industry signals such as C2PA or IPTC metadata, which can include images. For content that was only edited with AI, not generated, the label sits in the post's menu rather than on the post. Meta says it "may apply penalties" for failing to disclose without spelling them out, and may add a more prominent label to high-risk deceptive content. Meta restated these rules in February 2026, updated June 1, 2026.

Ads

There is no general AI-disclosure duty for ordinary commercial ads on Meta. For ads about social issues, elections or politics, advertisers must disclose during setup when the ad contains photorealistic images, video or realistic audio that was digitally created or altered to show a real person saying or doing something they did not, a realistic person or event that does not exist, or altered footage of a real event. Small edits like cropping, resizing and color correction are exempt. Undisclosed ads are rejected, and repeated failures can bring penalties. Since June 1, 2026, Meta also adds information to an "About this ad" panel for ads it detects were made with third-party AI tools. That happens automatically.

In the EU, Meta stopped delivering political, electoral and social-issue ads on October 6, 2025. Organic political posts are still allowed but cannot be boosted.

TikTok

Organic posts

TikTok's Community Guidelines require creators to label realistic AI-generated or significantly edited content (realistic scenes or people) with TikTok's AI-generated label. TikTok also labels content made with its own AI tools, and content carrying C2PA Content Credentials, automatically. Since November 2025 it has been testing invisible watermarks on AI content. Even labeled AI content can come down if it misleads on matters of public importance or harms people. Unlabeled or misleading AI content may be removed, made ineligible for the For You feed, or labeled by TikTok. We could not confirm every detail on TikTok's own guidelines page, and TikTok changes these rules often, so check the live Community Guidelines page before you post.

Ads

TikTok's ad policy (last updated April 2026) allows significantly edited or AI-generated content in ads only if it is disclosed, either with TikTok's AI-generated label or with your own clear caption, sticker, watermark or disclaimer. Minor touch-ups do not count. Separately, you may not use a public figure's likeness without permission, including to fake an endorsement, so an AI celebrity endorsement is out even with a label. Non-compliant ads are rejected or restricted.

Amazon Kindle Direct Publishing

If you self-publish books, workbooks or guides through KDP, the content guidelines require you to tell Amazon about AI-generated content (text, images including covers, or translations) when you publish a new title or edit and republish an existing one.

• AI-generated means created by an AI tool, even if you substantially edited it afterward. This must be disclosed.
• AI-assisted means you created the work yourself and used AI to edit, refine, error-check or otherwise improve it. This does not need disclosure.

Either way, you are responsible for making sure the content follows the guidelines. KDP can reject or remove titles that do not, and removed titles can be appealed. The disclosure is reportedly internal to Amazon and not shown to readers. In September 2023 KDP also capped new title creation at three per day. Reports of 2026 crackdowns with withheld royalties come from blogs, not Amazon, and we could not confirm them.

Etsy

Etsy's Creativity Standards allow items generated with AI from the seller's own original prompts under the "Designed by a seller" category, but the seller must say in the listing description that the item was created with AI. Selling AI prompt bundles is not allowed. Items made with computerized tools like laser cutters, 3D printers or Cricut machines must be based on your own original design, and production partners must be disclosed. Etsy can remove listings that do not comply.

This summary reflects Etsy's Creativity Standards as last updated June 10, 2025. Some seller blogs describe an August 2026 change that we have not been able to confirm on Etsy's own pages, so check the live page before you list.

LinkedIn

LinkedIn has no mandatory AI disclosure. When an image or video you post carries C2PA Content Credentials, LinkedIn shows a small icon that reveals whether AI generated or edited it, the app or device, and the signer. That also works for single-image and video Sponsored Content ads. In some countries LinkedIn lets members voluntarily mark uploads as AI-created. The feature is still rolling out. LinkedIn's help page.

There is no penalty for not disclosing, though misleading content can still be removed under LinkedIn's general community policies.

One habit that helps everywhere: keep your metadata

Many of these labels run on metadata embedded in the file: C2PA Content Credentials and IPTC source-type fields. AI tools add it; platforms read it. The weak link is usually in between. Image compressors, some content delivery networks, screenshot workflows and "save for web" exports can strip metadata without telling you. If you use AI images in product feeds, ads or posts, check that the metadata survives your workflow. In Merchant Center it is required. On YouTube, Meta, TikTok and LinkedIn it drives the automatic labels. And the EU now requires AI tool makers to embed machine-readable marks, as explained in the Article 50 guide.

The second habit: when a platform offers a disclosure toggle, use it, and add a plain caption too. The toggle satisfies the platform. The caption tells your customer. Wording ideas are on the examples page.