AI labels on ads and product images: Meta, TikTok and Google

Meta adds the label itself. Its help page says ads with images made or significantly edited with Meta’s own generative AI features, or made or edited with third-party AI tools, get an "AI info" label ^[2]. For outside tools, Meta says it detects them through industry-standard signals ^[1][3].

Where it shows: the label appears on the About this ad screen, opened from the three-dot menu at the top right of the ad, and sometimes also next to the Sponsored label at the top of the ad ^[2]. Meta’s 2025 announcement said that when its tools produce a photorealistic AI-generated person, the label goes next to Sponsored rather than behind the menu ^[1]. The 1 June 2026 update moved the labels into About this ad and added third-party detection; it does not repeat or withdraw the photorealistic-person rule ^[1]. On-model images of your garment are photorealistic AI people, so assume the label may be visible on the ad itself.

• Small changes such as resizing or colour correction generally don’t get the label ^[2].
• Meta says its own AI tools are rolling out gradually, so not every ad made with them carries the label yet ^[2].
• Ads about social issues, elections or politics must disclose AI use themselves ^[2][3]. The pages we read don’t describe a self-disclosure setting for other ads ^[2][3].
• Meta says the experience may vary by region because of legal requirements ^[1].

On TikTok the disclosure is your job. In Ads Manager, while you create the ad, you tick "This ad contains AI-generated content" and submit ^[4]. TikTok says to use it when all of these are true: it is a TikTok ad, it is not a Spark ad, and it includes an image, video or audio that is completely AI-generated or significantly modified by AI ^[4].

• Once submitted, you can’t untick it unless you change the image or video ^[4].
• New campaigns reset the setting, duplicates included, even when you reuse the same ad ^[4]. Tick it again each time.
• TikTok’s ad policy accepts its AIGC label or your own clear disclaimer, caption, watermark or sticker ^[5].
• The policy treats lighting, brightness and colour saturation changes, background removal or changes, and denoising as minor edits; content that is completely AI-generated counts as significant ^[5].
• If TikTok finds AI content that wasn’t disclosed, the ad is rejected or restricted; when you’re unsure, TikTok advises labelling it ^[5].

Spark Ads, which boost an organic post, follow the Community Guidelines instead ^[4]. Those ask creators to label AI-generated or significantly edited content that shows realistic-looking scenes or people, and say unlabelled content may be removed, restricted or labelled by TikTok ^[6]. TikTok has also read C2PA Content Credentials since May 2024 to label AI images and videos made on other platforms automatically ^[7].

A new style of your garment on an AI model is a completely AI-generated image, even though the garment design comes from your bestseller. Tick the setting.

In July 2026 Google started letting advertisers label image and video ads made or edited with AI, either with a label added to the creative or with a new AI label setting, rolled out through July across Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center and Ads Editor ^[8]. Google says these labels don’t count against its text overlay and watermark rules, and that the change helps with transparency rules in the EU, India and New York ^[8].

• You label each asset, in the asset library or when you review assets while creating a campaign ^[9].
• Labelled ads show a note in the "How this ad was made" part of My Ad Center, reached from the ad’s three-dot menu ^[9].
• Campaigns aimed at the EU, India or New York also get a visible label on the ad itself ^[9].
• Google may add labels itself, for example where a law requires it or when it receives signals from other platforms, and you can’t remove those ^[9].
• Election ads still use the separate altered or synthetic content checkbox ^[8].

Google says using the setting doesn’t guarantee you meet any specific law and that you should get legal advice ^[8][9].

Shopping listings work through the image file, not a tick box. Merchant Center says every AI-generated image must carry the IPTC DigitalSourceType tag trainedAlgorithmicMedia, and that you must not strip it ^[10][11]. This covers the main image, additional images and lifestyle images (image_link, additional_image_link and lifestyle_image_link) ^[10][11].

IPTC, the body that maintains the standard, defines trainedAlgorithmicMedia as media made by an AI model trained on captured content, labelled "Created using Generative AI" ^[12]. Merchant Center also names compositeSynthetic, for images that mix in generative elements, and algorithmicMedia as values to keep ^[10][12].

• Re-saving, compressing or resizing an image in some tools can drop its metadata. Check the tag is still there in the file you actually upload.
• If AI wrote a product title or description, Merchant Center asks you to send it in structured_title or structured_description, with digital_source_type set to trained_algorithmic_media ^[10].

Article 50 of the EU AI Act has applied since 2 August 2026 ^[13]. It splits the work in two. Providers, the companies that make AI tools, must mark generated images in a machine-readable way so they can be detected as AI-made ^[13]. Deployers, meaning any business using an AI system in its work, must clearly disclose deepfakes ^[13]. The European Commission’s FAQ gives an advertising company as an example of a deployer ^[13], so a clothing brand using AI images in its marketing is likely one too.

The FAQ defines a deepfake as AI-generated or manipulated image, audio or video that resembles existing persons, objects, places or events and would falsely appear authentic, and it says the audience and the context matter ^[13]. It doesn’t mention product images or ads ^[13]. Whether an AI model wearing a style you haven’t made yet counts is not settled, which is one more reason to say plainly that it’s a concept image.

• Where a deepfake disclosure is needed, it must be visible or audible at the latest when someone first sees the content; the provider’s hidden marking alone is not enough ^[13].
• Tools already on the market before 2 August 2026 have until 2 December 2026 for the machine-readable marking only ^[13].
• Content made before 2 August 2026 doesn’t have to be labelled after the fact, though the Commission encourages it ^[13].
• A voluntary Code of Practice on marking and labelling AI-generated content exists for providers and deployers ^[13].

This is our wording, not any platform’s. Turning on a platform’s own setting comes first; a line in the ad text or on the product page is on top of it. Edit it to match what you actually did.

Keep the ship date real. Our guide to pre-orders for clothing covers dates and delays. Don’t write that an image is approved, compliant or free of AI labels; you can’t promise what a platform will do.

• 9 May 2024: TikTok starts reading C2PA Content Credentials to label AI images and videos made on other platforms ^[7].
• 3 February 2025: Meta announces AI info labels on ads made or significantly edited with its own AI tools, with photorealistic AI people labelled next to Sponsored ^[1].
• September 2025: last update of TikTok’s Ads Manager help page on the AI-generated content setting ^[4].
• April 2026: last update of TikTok’s ad policy on misleading and false content, which sets the AI disclosure rules ^[5].
• 1 June 2026: Meta moves AI info into About this ad and starts detecting ads made with third-party AI tools ^[1][3].
• 9 July 2026: Google announces AI labels for ads and the AI label setting, rolled out through July ^[8].
• 2 August 2026: EU AI Act Article 50 applies ^[13].
• 24 September 2026: the current TikTok Community Guidelines, including the AI content rules, take effect ^[6].
• 2 December 2026: end of the grace period for machine-readable marking by AI tools already on the EU market ^[13].

Will an AI label hurt my ad’s results?

None of the platforms publish figures on this, and we haven’t seen reliable ones. If you’re worried, test it: run the labelled concept images against each other, not against an unlabelled photo, so the label doesn’t decide the result.

Can I remove the metadata so no label appears?

Don’t. Merchant Center tells you not to strip it ^[11], TikTok rejects or restricts ads with undisclosed AI content ^[5], and Meta can label ads on other signals too ^[1]. Hiding it also misleads the buyers you are trying to learn from.

The garment is my real bestseller. Only the model and the new colour are AI. Is the image still AI-generated?

Yes. The whole image is generated, so it counts as completely AI-generated for TikTok ^[4][5] and needs the trainedAlgorithmicMedia tag for Merchant Center ^[10]. Say that the design is yours and the image is a concept.

I only resized my own photo or corrected the colour. Do I need a label?

Meta and TikTok both treat small edits like these as minor ^[2][5]. A real photo of your real garment, lightly edited, is not what these rules are about.

Do I need to say anything on my Shopify product page?

We didn’t find a Shopify rule on this. Buyers can’t tell a concept image from a photo of stock, though, so say it next to the images, and swap in real photos once the piece is made.