Google Merchant Center AI Image Metadata Requirements (2026)

Why Google Merchant Center AI image rules matter in 2026

Google Shopping and free listings still run on product feeds — and feeds still live or die on image quality. The twist in 2026 is not that AI product photos are banned. It is the opposite: Google Merchant Center allows generative-AI images in Shopping attributes when they accurately depict the product and retain the right IPTC provenance metadata.

That combination catches many sellers off guard. Teams generate clean packshots, run them through a CDN or compression pipeline, strip EXIF/IPTC tags for "leaner" files, and then wonder why Merchant Center warnings appear — or why an AI workflow that looked fine on Amazon fails a Google crawl. Separately, Google is raising the minimum resolution for image_link and additional_image_link to 500×500 with warnings already rolling out and hard enforcement on January 31, 2027.

This guide is a practical checklist drawn from Google Merchant Center Help concepts (AI-generated content, image link attributes, and the 2026 product data specification update). It is not a substitute for Google's official Help Center, but it maps the rules sellers actually break — and how to stay feed-compliant while still shipping AI photography at scale.

If you already maintain Shopping feeds, pair this with our broader Google Shopping product image requirements AI guide and compare channel norms against Amazon product image requirements.

AI images are allowed — with accuracy and metadata

Which feed attributes accept AI images

Per Google's AI-generated content guidance, generative-AI images may be used in:

• Image link [image_link] — your primary product image
• Additional image link [additional_image_link] — secondary angles, details, and staging
• Lifestyle image link [lifestyle_image_link] — in-context / lifestyle shots

The policy hinge is simple: the image must accurately depict the product you are selling. AI that invents a different colourway, adds missing accessories, morphs logos, or "improves" packaging into something you do not ship is a misrepresentation risk — metadata will not save a false product photo.

Accuracy first, then disclosure

Treat AI as a production tool, not a product designer. Start from a real photograph of the SKU you fulfil. Generate white-background packshots, alternate angles, or lifestyle scenes that keep shape, colour, branding, and materials faithful. Then QA against the physical sample before the URL ever hits your feed.

That workflow mirrors how careful sellers already approach TikTok Shop image requirements and Amazon galleries: channel rules differ, but truthfulness is universal.

IPTC DigitalSourceType: the metadata you must not strip

What Google requires

All images created with generative AI must include metadata indicating AI origin via the IPTC DigitalSourceType property. Google calls out preserving tags such as:

• TrainedAlgorithmicMedia — image created using a model derived from sampled training content (the common generative-AI case)
• CompositeSynthetic — a composite that includes synthetic elements
• AlgorithmicMedia — created purely by an algorithm not based on sampled training data

Google's Help Center is explicit: do not remove embedded IPTC DigitalSourceType metadata from images created with generative AI tools (Product Studio is cited as an example). The disclosure lives in the image file that Google crawls from your URL — not as a free-text note in your feed spreadsheet.

Why tags disappear in real pipelines

Most accidental non-compliance is operational, not intentional:

1. 1.Export settings that "remove metadata" or "strip EXIF" for privacy/size
2. 2.CDN transforms that recompress or re-encode and drop XMP/IPTC blocks
3. 3.Screenshot / flat re-save workflows that bake pixels but lose provenance
4. 4.Third-party optimizers that treat metadata as optional baggage
5. 5.Manual editor saves without preserving XMP

If your AI tool embeds TrainedAlgorithmicMedia correctly but your hosting stack rewrites the file, Merchant Center sees a generative-looking image without the required disclosure path Google expects.

Practical preservation checklist

• Confirm your generator writes IPTC/XMP DigitalSourceType on export
• After CDN upload, download the exact URL in the feed and inspect metadata (not only the local master)
• Prefer stable, crawlable object URLs over aggressive on-the-fly rewrite endpoints
• Avoid "optimize and strip" presets for Shopping assets; optimize dimensions/quality without nuking XMP
• Document the pipeline so every new SKU inherits the same preserve-metadata path

When in doubt, treat DigitalSourceType like a compliance label on the binary — same seriousness as accurate GTIN data.

Core image_link rules that still apply to AI shots

AI permission does not relax classic image_link quality rules. Google still expects images that:

• Accurately display the product
• Use supported formats such as JPEG, WebP, PNG, non-animated GIF, BMP, or TIFF
• Are reachable via crawlable HTTP/HTTPS URLs (encoded per URL standards; not blocked from Googlebot / Googlebot-image in robots.txt)
• Avoid promotional text, watermarks, borders, and placeholders / generic images
• Are not merely upscaled thumbnails

Those constraints apply whether the pixels came from a studio strobe or a diffusion model. An AI packshot with a sale banner, a stock watermark, or a grey "image coming soon" tile fails for the same reasons a human photo would.

Main vs additional vs lifestyle

Use [image_link] for the clearest primary representation of the SKU. Use [additional_image_link] for secondary angles, detail crops, and staging that still meet image requirements. Use [lifestyle_image_link] when you have a true in-context shot — and keep IPTC AI tags intact there too.

For sizing strategy across marketplaces, see best product photo sizes for Shopify, Amazon, and Google.

Size requirements: 1500×1500 recommended, 500×500 deadline

What is changing

Google announced updated minimum resolution requirements for images submitted in image_link and additional_image_link:

• New minimum: 500 × 500 pixels across product categories and marketing methods
• Warnings: Merchant Center began issuing warnings for undersized images starting April 14, 2026 (look for messaging such as "Image too small for upcoming enforcement" under Needs attention)
• Enforcement: January 31, 2027

Until enforcement, legacy floors (for example smaller non-apparel / apparel minima depending on context) may still apply for some products, but waiting is a poor strategy — especially if Google also auto-optimizes some small images and flags them for review.

What Google recommends for performance

Even with a 500×500 floor, Google recommends providing images around 1500 × 1500 pixels or larger for best performance across listing formats. Also respect upper bounds called out in Help (for example very large megapixel counts and file-size caps such as 16MB).

Action plan before January 2027

1. 1.Audit feed URLs under Needs attention for size warnings
2. 2.Re-export masters at 1500×1500+ where possible (square often travels best across surfaces)
3. 3.Replace soft or upscaled thumbnails — do not just stretch tiny assets
4. 4.Re-check IPTC tags after re-export and CDN publish
5. 5.Monitor again after feed fetch cycles so crawls pick up the new binaries

A feed-compliant AI photography workflow with Sellable

The safest pattern for Merchant Center is boring on purpose:

1. 1.Capture or select a real product photo of the exact SKU
2. 2.Generate accurate white-background packshots and optional lifestyle frames with AI that stays faithful to the reference
3. 3.Export at Shopping-ready resolution (aim 1500×1500+) in a supported format
4. 4.Preserve IPTC DigitalSourceType through hosting
5. 5.QA colour, branding, and accessories against the physical product
6. 6.Publish crawlable HTTPS URLs into image_link / additional_image_link / lifestyle_image_link

Sellable is built for that loop: start from a real product photo, generate packshots and lifestyle creatives that stay visually accurate, and keep outputs suitable for multi-channel feeds — including Google Shopping workflows where metadata preservation and clean, non-promotional frames matter.

You do not need a different "Google-only" creative language. You need accurate pixels, intact AI provenance tags, and URLs that survive crawl + policy checks. Soft CTA, hard requirement: if your current toolstrip strips IPTC on upload, fix hosting before you scale generation.

Common failure modes (and quick fixes)

Checklist: Merchant Center AI image compliance 2026

• [ ] AI used only where the result accurately depicts the sellable SKU
• [ ] Images used in image_link, additional_image_link, and/or lifestyle_image_link as appropriate
• [ ] IPTC DigitalSourceType present (TrainedAlgorithmicMedia, CompositeSynthetic, or AlgorithmicMedia as applicable)
• [ ] Metadata still present on the live crawled URL
• [ ] No promo text, watermarks, borders, or placeholders
• [ ] Supported format; crawlable HTTPS URL
• [ ] Resolution on a path to 500×500 minimum by January 31, 2027 enforcement — preferably 1500×1500+ now
• [ ] Feed re-fetched after URL updates; Needs attention reviewed

FAQ

Does Google ban AI product photos in Merchant Center?

No. Google allows generative-AI images in image_link, additional_image_link, and lifestyle_image_link when they accurately show the product and retain required IPTC DigitalSourceType metadata.

Is DigitalSourceType a feed column?

For images, disclosure is embedded in the image metadata Google retrieves from your image URL. (Separately, AI titles and descriptions use structured attributes with a digital_source_type sub-attribute — do not confuse that text workflow with image IPTC tags.)

What size should I upload today?

Upload 1500×1500+ when you can. Treat 500×500 as the upcoming hard floor (warnings from April 14, 2026; enforcement January 31, 2027), not as your quality target.

Can I add a sale badge to an AI packshot?

Avoid promotional text overlays, watermarks, and borders on images you submit for these attributes — AI or not.

Next step

Audit one SKU end-to-end: open the live image_link URL, confirm crawlability, inspect IPTC DigitalSourceType, check pixel dimensions, and remove any overlay debt. Then roll the same pipeline across the catalog.

When you want accurate packshots and lifestyle frames from a real product photo without fighting feed rules, try Sellable and keep your Merchant Center images both beautiful and policy-aligned.