Do you have to label AI-generated content on Instagram? 2026 rules by platform

The platforms and the EU draw roughly the same line: a realistic picture, clip or voice that shows something that did not happen needs a label, and a colour grade does not. The details differ by platform, and the EU rule sits on top of all of them. This page puts each platform's rule in one table, sets the EU AI Act's transparency duties beside it, and ends with a decision list a brand can run on every post. Every rule below was read on the platform's or the EU's own pages on 2026-09-28.

Do you have to label AI-generated content on Instagram?

For video and audio, yes. Meta requires people who post organic content with photorealistic video or realistic-sounding audio that was digitally created or altered to use its AI disclosure tool, and says it may apply penalties when they do not (Meta, February 2024). Meta restated the requirement in its February 2026 elections update. The rule covers Instagram, Facebook and Threads. Viewers see the label as "AI info"; Meta renamed it from "Made with AI" on 1 July 2024 (Meta, April 2024, with dated updates).

Still images are not named in the self-disclosure duty. Meta labels them itself when it detects industry-standard signals: "AI generated" information in C2PA and IPTC metadata, and invisible watermarks. Since 12 September 2024, when Meta detects that content was only edited with AI tools rather than generated, it moves the "AI info" label into the post's menu. Where content carries a particularly high risk of materially deceiving the public on a matter of importance, Meta says it may add a more prominent label.

Two more Meta rules reach some brands. Advertisers must disclose AI use in certain ads about social issues, elections or politics, and Meta adds that disclosure to the ad and to the Ad Library (Meta, 2026). And in a post dated 31 August 2026, Instagram renamed its "AI creator" label, which it began testing earlier in 2026, to "AI-generated profile": a label for profiles that feature an AI-generated person instead of a human. It says profiles that add the label keep their reach, unlabelled ones may see limits, and creators who only use AI tools in their process do not need it (Instagram for Creators). A brand with a virtual ambassador, or an anonymous account built around an AI persona, is in the first group.

What does each platform require?

Checked on each platform's own pages on 2026-09-28. "Not stated" means the page cited does not say.

Platform	What you must label	How you label it	What the platform labels by itself	If you don't
Instagram, Facebook, Threads	Organic posts with photorealistic video or realistic-sounding audio that was digitally created or altered	Meta's AI disclosure tool when posting; viewers see "AI info"	Content with AI signals in C2PA or IPTC metadata or invisible watermarks; edited-only content gets the label in the post menu	Meta "may apply penalties"; a more prominent label for high-risk deception
TikTok	AI-generated or significantly edited content that shows realistic-looking people or scenes	TikTok's AIGC label, or your own clear caption, sticker or watermark	Content made with TikTok's AI effects; uploads carrying C2PA Content Credentials	Removed, restricted or labelled by TikTok, depending on the harm
TikTok ads	Ads with content significantly modified by AI, meaning changed beyond minor tweaks	AIGC label, or a clear disclaimer, caption, watermark or sticker	Not stated	Ad rejected or restricted
YouTube	Realistic content that makes a real person appear to say or do something they didn't, alters footage of a real event or place, or shows a realistic scene that didn't happen	YouTube Studio upload flow: Attributes, "AI use", Yes	Content made with YouTube's AI tools, content with C2PA metadata, and content its systems detect	For creators who consistently don't disclose: a label applied manually, or penalties including content removal or suspension from the YouTube Partner Program
LinkedIn	Synthetic or manipulated media showing a person saying or doing something they did not, unless the fake or altered nature is clearly disclosed	The LinkedIn pages cited here describe no labelling setting, so the disclosure goes in the post text or on the media	A Content Credentials icon on images and video that carry signed C2PA data (rolling out gradually)	A breach of LinkedIn's Professional Community Policies

Sources for the rows: TikTok Community Guidelines (released 25 August 2026), TikTok's Content Credentials announcement (9 May 2024), TikTok's ads policy on misleading and false content, YouTube Help, LinkedIn's Professional Community Policies and LinkedIn Help on Content Credentials.

Three details from those pages change what a brand should do:

Where the label shows. On YouTube, photorealistic AI content may get a label in the video player; non-photorealistic or animated content may get it in the expanded description. On Meta, content detected as only edited gets the label in the menu. A label in a menu or a description is easy to miss.
Metadata can label a post for you. TikTok, YouTube, Meta and LinkedIn all read C2PA data. If your design or editing tool writes Content Credentials into an export, a post you consider a minor edit can arrive labelled. On YouTube, a label that came from C2PA metadata cannot be switched off by the creator.
Disclosing does not cost reach on YouTube. YouTube says disclosing AI content won't limit a video's audience or its eligibility to earn money.

Which edits count as minor?

YouTube and TikTok publish examples on both sides of the line. Meta publishes a test (photorealistic, digitally created or altered) and names one minor modification, AI retouching tools, in its July 2024 update. The table lists only what the two pages state.

Edit	YouTube	TikTok
Colour or lighting correction	No label	No label for colour correction; lighting not listed
Cropping or reframing	Not listed	No label
Beauty filter; background blur or vintage effect	No label	Not listed
AI help with a script, title, thumbnail or captions	No label	Not listed
Generic text-to-speech narration	Not listed	No label, unless it is a recognisable voice of a known person
Cloning your own voice for a voice-over or dub	No label	Not listed; AI audio that mimics a real person's voice needs a label
Cartoon, anime or clearly fantastical content	No label	No label for artistic styles such as anime
AI generating or extending a backdrop, such as to simulate a moving car	No label	Not listed
Adding or removing a background, object or person	Label if it alters footage of a real event or place	Label if done in a misleading way
AI-generated footage of a real place for a promotional video	Label (YouTube's own example)	Label (realistic-looking scene)
A real person saying or doing something they didn't	Label	Label
AI-generated music	Label	Not listed

YouTube's example of a disclosable video is a marketing one: AI-generated extra footage of a real place, "like a video of a surfer in Maui for a promotional travel video".

What does the EU AI Act require from 2 August 2026?

Article 50 of the AI Act, Regulation (EU) 2024/1689, sets transparency duties that apply from 2 August 2026, the regulation's general date of application under Article 113. The Commission's Article 50 FAQ (updated 24 July 2026) confirms the date. For a business posting on social media, two duties matter, both in Article 50(4) and both placed on the "deployer".

Who counts as the deployer?

A deployer is any person or organisation using an AI system under its authority, except in "a personal non-professional activity" (Article 3(4)). The Commission's FAQ gives an advertising company as the example: the company is the deployer, not the employees who work under its instructions, such as web designers or content creators, and it stays the deployer when contractors or freelancers are involved. A freelancer who gains an economic benefit on a regular basis is acting professionally and is a deployer in their own right.

Location does not take a business out of scope. The Act covers deployers established in the EU, and deployers outside the EU where the AI system's output is used in the Union (Article 2(1)(b) and (c)).

Deepfakes: a disclosure people can see or hear

Article 3(60) defines a deep fake as AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. Note the list: objects and places count, not only people. The Commission's FAQ splits this into three cumulative criteria: resemblance, a subject that exists or could plausibly exist, and a false appearance of authenticity. For the last one, deployers may weigh the level of resemblance, the message, the context and what the intended audience expects.

A deployer must disclose that such content was artificially generated or manipulated. The disclosure must be clear and distinguishable, at the latest at first exposure, and meet accessibility requirements (Article 50(5)). The Commission adds that it must be perceivable without special tools, for example a visible or audible label, and that deployers cannot rely solely on the machine-readable marking that AI providers embed.

For evidently artistic, creative, satirical or fictional work, the duty shrinks to disclosing the generated content in a way that does not hamper the display or enjoyment of the work.

AI text on matters of public interest

The second duty covers AI-generated or manipulated text published to inform the public on matters of public interest. The Commission's list of such matters is broad: politics, public health, environmental protection and consumer safety are on it, as are economic, financial, scientific or cultural developments that may be the subject of public debate. Text is exempt when it has gone through human review or editorial control and a person or company holds editorial responsibility. Spell-checking or grammar fixes do not count as review.

The EU icons

The Commission published a set of EU icons for labelling AI-generated content (page updated 24 September 2026). There are three types, a basic icon, "fully AI-generated" and "partially AI-modified", each in four versions: black, white, and both at 50% transparency. Using them is optional; the labelling duty is not, and using the icon does not by itself establish compliance. The display rules summarised on that page: visible at first exposure, not covered by overlays, embedded in the content itself (creative works excepted) unless an equivalent alternative such as a user-interface overlay is available, and still visible when the content is reshared or downloaded.

One of the Commission's examples of a partially AI-modified deepfake is a brand scenario: authentic photographs of an empty apartment, furnished using AI.

The icons belong to the Code of Practice on transparency of AI-generated content, published in final form on 10 June 2026. The Commission and the AI Board have assessed it as adequate for showing compliance. Signing it is voluntary, and deployers can sign as well as providers.

What the Digital Omnibus changed, and what it did not

Regulation (EU) 2026/1744, the Digital Omnibus on AI, is dated 8 July 2026, was published in the Official Journal on 24 July 2026 and entered into force on the third day after publication. For Article 50 it did two things. It added Article 111(4): providers of generative AI systems placed on the market before 2 August 2026 must comply with the Article 50(2) marking duty by 2 December 2026. And it rewrote Article 50(7) on how codes of practice are assessed. It did not amend Article 50(4).

Duty	On whom	Applies from
Disclosure of deepfakes, Art. 50(4)	Deployers: brands, agencies, freelancers acting professionally	2 August 2026
Label on AI text about matters of public interest, Art. 50(4)	Deployers	2 August 2026
Clear, accessible, at first exposure, Art. 50(5)	Providers and deployers	2 August 2026
Machine-readable marking of AI output, Art. 50(2)	Providers of generative AI systems	2 August 2026; 2 December 2026 for systems already on the market before 2 August 2026

The grace period belongs to the AI tool vendors. Brands got none.

Fines

Breaching Article 50 can draw an administrative fine of up to EUR 15 million or 3% of total worldwide annual turnover, whichever is higher (Article 99(4)(g)). For SMEs, and after the Omnibus also for small mid-caps, the cap is whichever of the two is lower (Article 99(6) and (6a)). Member States set the penalty rules and enforce them.

Is a platform's AI label enough for the EU rule?

None of the Commission or platform pages cited here says so. Our reading of the rules above is that a platform label may fall short for a deepfake shown to people in the EU, for three reasons. Meta moves the label into a menu for content it detects as edited. YouTube may place it in the expanded description. And a platform label does not travel with a download or a repost on another network, while the Commission's icon guidance asks for a label embedded in the content, or an equivalent alternative such as a user-interface overlay, that stays visible when the content is reshared or downloaded.

The conservative setup for EU-facing deepfake content is both: the platform's own setting, and a short on-screen or in-image disclosure in the first frame or corner, such as "AI-generated" or the EU icon with a text label.

A decision list for every brand post

Steps 1 to 4 decide whether a post needs an AI label; stop at the first one that answers it. Steps 5 to 8 apply to every post that used AI.

Was any part of the image, video or audio made or changed by AI? If no, no AI label. AI-written text is covered in step 7.
Was the change only a standard edit: colour, lighting, crop, reframe, blur, sharpening, upscaling or audio repair? Between them, YouTube and TikTok list these as edits that need no label; Meta's pages name only AI retouching as a minor modification. Check whether your export tool wrote C2PA data, because a platform may label the post anyway.
Is the result clearly unrealistic: a cartoon, an anime style, an obvious fantasy? YouTube and TikTok ask for no label. Under the EU definition, content that would not falsely appear authentic to its audience is not a deepfake.
Does it show a real person saying or doing something they didn't, a real place or event altered, or a realistic scene that never happened? Label it on every platform: Meta's tool for video and audio, TikTok's AIGC label, YouTube's "AI use", a written disclosure on LinkedIn. If the audience includes people in the EU, add a visible disclosure in the content itself.
Does it use a real person's face or voice? Get written consent first. A label does not cure a missing consent: TikTok does not allow the likeness of private individuals without consent even when labelled, and its ads policy bars misuse of a public figure's likeness without their permission.
Is it a paid ad? On TikTok, disclose any content significantly modified by AI or the ad is rejected or restricted. On Meta, ads about social issues, elections or politics have a separate AI disclosure.
Is it text that informs the public on a matter of public interest, such as a statement on a product safety issue? If AI wrote it and no one with editorial responsibility reviewed it in substance, the EU rule requires a label. A real review removes the duty.
Then check the post's other rules. A paid partnership needs its own #ad disclosure, and a giveaway needs official rules and a platform release.

Six brand posts run through the list

The EU column is our reading of Article 50 and the Commission's guidance, not a ruling.

Post	Platforms	EU AI Act	What to do
A bakery Reel with an AI colour grade	Minor edit on YouTube and TikTok; Meta lists no colour-grade example	No false appearance of authenticity	No label on YouTube or TikTok; on Instagram, the one judgement call in this table
Listing photos of an empty flat, furnished with AI	TikTok: object added in a misleading way, label. Instagram: still image, not in Meta's self-disclosure duty, may be auto-labelled	The Commission's own example of partially AI-modified content	Label in the image and in the caption
A promo video with AI footage of a real beach the brand never filmed	YouTube's own example; realistic scene on TikTok; photorealistic video on Meta	Resembles an existing place and would pass as real	Platform setting on every network, plus an on-screen label
An ad voice-over generated to sound like a known actor	TikTok: AI voice of a real person needs a label, and the ads policy bars misuse of a public figure's likeness without their permission. Meta: realistic-sounding audio	Audio resembling an existing person	Do not run it without the actor's written permission; with it, label on every platform and in the ad
An AI-illustrated cartoon mascot	No label on YouTube or TikTok; Meta may auto-label from metadata	Not presented as authentic	No label needed
A product-launch caption drafted with AI and edited by staff	None of the platform rules above covers caption text	Not public-interest text, and it was reviewed	No label; keep the review step real

Frequently asked questions

Does an AI label reduce a post's reach?

YouTube says disclosing AI content won't limit a video's audience or its eligibility to earn money. Instagram says profiles that add its AI-generated profile label keep their reach, while unlabelled profiles featuring an AI-generated person may see limits. The Meta and TikTok pages cited here say nothing about the reach of individual labelled posts.

Do I have to label AI-generated images on Instagram?

Meta's self-disclosure duty names photorealistic video and realistic-sounding audio, not still images. Meta labels images itself when it detects AI signals such as C2PA or IPTC metadata. The same image may still need a label elsewhere: on TikTok if it shows a realistic-looking scene or person, and under the EU AI Act if it is a deepfake shown to people in the EU.

Does the EU AI Act apply to a small business or a freelancer?

Yes, when AI is used for work. Only personal, non-professional use falls outside the deployer definition, and the Commission treats a freelancer with a regular economic benefit as a deployer. Businesses outside the EU are covered when their AI output is used in the EU. For SMEs, the fine cap is the lower of EUR 15 million and 3% of turnover.

Does the 2 December 2026 grace period apply to brands?

No. Article 111(4), added by Regulation (EU) 2026/1744, gives only providers of generative AI systems already on the market until 2 December 2026, and only for machine-readable marking under Article 50(2). The deployer's deepfake and text duties in Article 50(4) apply from 2 August 2026.

Can I remove a label a platform added automatically?

On YouTube, in most cases: select No under "AI use" in the Attributes section. That does not work for content made with YouTube's AI tools, content with C2PA metadata, or content labelled after a manual review. The Meta, TikTok and LinkedIn pages cited here do not describe a removal route.

Sources

All read on 2026-09-28.

Regulation (EU) 2024/1689 (AI Act), Articles 2, 3(4), 3(60), 50, 99, 111 and 113, Official Journal text. eur-lex.europa.eu
Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal L, 24 July 2026: amendments to Articles 50(7), 99 and 111(4). eur-lex.europa.eu
European Commission, "Transparency obligations under Article 50 AI Act" FAQ, updated 24 July 2026. digital-strategy.ec.europa.eu
European Commission, Code of Practice on Transparency of AI-generated Content, updated 31 July 2026. digital-strategy.ec.europa.eu
European Commission, EU icons for labelling AI-generated content, updated 24 September 2026. digital-strategy.ec.europa.eu
Meta, "Labeling AI-Generated Images on Facebook, Instagram and Threads", 6 February 2024. about.fb.com
Meta, "Our Approach to Labeling AI-Generated Content and Manipulated Media", 5 April 2024, updates of 1 July and 12 September 2024. about.fb.com
Meta, "How Meta Is Preparing for the 2026 US Midterm Elections", 19 February 2026, updated 1 June 2026. about.fb.com
Instagram for Creators, "Bringing More Transparency to AI-Generated Profiles on Instagram", 31 August 2026. creators.instagram.com
TikTok Community Guidelines, Integrity and authenticity: edited media and AI-generated content, released 25 August 2026. tiktok.com
TikTok Newsroom, "Partnering with our industry to advance AI transparency and literacy", 9 May 2024. newsroom.tiktok.com
TikTok Advertising Policies, "Misleading and false content". ads.tiktok.com
YouTube Help, "Disclosing use of GenAI content". support.google.com
LinkedIn Help, "Content credentials". linkedin.com
LinkedIn Professional Community Policies. linkedin.com