Meta catalog ads dynamic creative: what the feed owns

Performance marketers reach for the phrase "dynamic creative" and assume it names a setting on a Meta catalog ad: a toggle that hands Meta your products and lets it design the image. It does not. On Meta Advantage+ catalog ads (the format that used to be called Dynamic Product Ads, or DPA), the thing Meta varies dynamically is which product it shows which person. The picture starts as whatever URL sits in your feed. Meta fetches that URL, caches the pixels, and reuses them. Meta's own image tools, such as image templates and dynamic overlays, build on that photo, and they are covered below.

How Advantage+ catalog ads actually work

You upload a product catalog to Meta Commerce Manager, fed by a data file (CSV, TSV, or XML) or a platform connector. You build one ad. Meta populates it per impression with the right product's image, title, price, and link pulled from the catalog, and decides which product to show which user. In 2026 this is the default recommendation for most e-commerce advertisers: a single campaign that blends prospecting and retargeting and auto-allocates budget across Facebook, Instagram, Audience Network, and Messenger.

The matching is where the "dynamic" reputation comes from, and it is product matching, not creative generation:

• Retargeting: Meta matches catalog products to a user's on-site behaviour via the Meta Pixel (or the Conversions API). The standard events are ViewContent, AddToCart, and Purchase, each passing content_ids that must exactly match the catalog product id (with content_type=product). Mismatched IDs mean no match, which means no dynamic retargeting, silently.
• Prospecting: Meta predicts which people in a broad audience are most likely to want catalog products, from on- and off-Facebook behaviour, with no prior site visit required.

Everything Meta assembles in the ad (image, title, price, link) is read from your feed fields. Nine fields are required for a Meta catalog feed:

For the full field reference and image dimensions, see the Facebook catalog image specs page.

The image is just a URL (the load-bearing fact)

Here is the single most important mechanical truth for anyone trying to improve catalog-ad creative: image_link is a plain URL. Meta fetches that URL, caches the image, and uses the cached pixels in the dynamically assembled ad. Meta does not tell a pre-rendered design at that URL apart from a raw photo: whatever sits there is the image the ad starts from.

The consequence is the entire thesis behind feed-image tools: if you put a pre-rendered, branded image URL in image_link, Meta serves it like any other photo. No special API, no "dynamic creative" integration, no flag to set. This is exactly how Emberfeed works. It rewrites g:image_link in the served feed to point at its on-demand render endpoint, and Meta caches that render the same way it would cache a CDN photo. It is also how the broader category (Confect, VersaFeed, Socioh's "Branded Catalogs", and others) operates.

What Meta's own creative tools do

Meta does give advertisers native creative controls on catalog ads, and any honest treatment has to say so. If they cover your need, use them: they are free and built in.

Dynamic overlays (the feed-data stickers)

Meta natively renders sticker-style labels pulled straight from catalog data: current price, strikethrough sale price, percentage off, and free shipping. You switch them on in Ads Manager, in the ad's Advantage+ creative settings. As of the April 2025 update you can toggle each label independently (previously you got one at a time), pick the shape (rectangle, pill, circle or triangle), colours and font, or let Meta add the most relevant ones automatically. The free-shipping label requires setting the item's shipping cost to 0 in the catalog.

The catch worth knowing: Meta's help page says dynamic overlays only show on Facebook Feed, not on Instagram, Stories or Reels. So they are a Facebook Feed tool, not an everywhere tool.

Image templates (and the end of frames)

Image templates are Meta's newer and much bigger tool. In Commerce Manager you build a layout from layers: the product photo (with optional background removal), catalog fields such as title, price and sale price, your own text, a logo or other image, stickers and a background colour. Catalog field, image and text layers can be set to show only on products that meet a rule on a product attribute, such as price, so one template can put different content on different products. The design starts at 1:1, with optional 4:5 (Feed) and 9:16 (Stories and Reels) versions, and you choose the template per ad in Ads Manager. Templates work with single image and carousel ads, not collection.

Image templates replace the old frame, one transparent 1:1 PNG laid over every product in the ad. Meta made October 30, 2026 the last day frames are available, so a frame design has to be rebuilt as a template. We compare templates with a pre-rendered feed image in detail in Meta image templates vs Emberfeed.

Where Meta's native tools stop

The native tools are genuinely useful, and since image templates they cover far more than a price sticker. The limits that remain follow from where the design lives: Meta applies it when it builds the ad, so it exists only inside Meta ads.

So the honest rule is simple. If your catalog only runs on Meta and the design is a price label, a sale price, your logo and some text, Meta's dynamic overlays and image templates do it for free, and you should use them. You need a real image in the feed when the design has to go beyond Meta, to other platforms that read your feed such as TikTok; when it needs logic or layouts a layer editor cannot express; or when it has to show in formats templates skip, such as collection ads. That is the gap a feed-image renderer fills. Emberfeed's specific wedge on top of that is AI-assisted template design (the editor composes a brief for Claude or ChatGPT, or Claude builds the template over Emberfeed's MCP connector), roughly 25 €/feed pricing, and single-tool simplicity instead of a multi-product feed stack. It is not self-hosted, and it needs your source feed to work: it renders templated images per product from a feed you import, it does not invent a catalog.

Per-placement reality: square is the safe default

Because a pre-rendered image travels to every placement, you have to respect each placement's shape. The 2026 consensus:

• Feed: 1:1 (square) or 4:5.
• Stories: 9:16.
• Reels: 9:16 (1080×1920) is the dominant, safe spec. One vendor reports Meta serving 4:5 in some Reels contexts as of mid-2025, so treat 9:16 as the safe default and do not assume 4:5 is settled.

For a 9:16 canvas (1080×1920), keep the top and bottom roughly 250 px each clear of logos, price, and CTA. Meta's own UI (profile, CTA button, navigation, the Reels right-rail buttons) covers those zones. Practitioners use a rule of thumb of about 14% off the top and 20–35% off the bottom, with the load-bearing information kept inside a centred 1:1 box.

Creative variables that move catalog-ad CTR

A note on the numbers below before you read them: every lift figure in this section comes from feed-tool and agency blogs, companies that sell the outcome. None is a Meta-published study or a controlled experiment. Treat the variables as the durable, defensible takeaway and the percentages as illustrative, vendor-reported, single-brand anecdotes. They tell you what to test, not what you will get.

Strip the percentages away and the durable lessons hold up across every source:

1. Lifestyle and context beat a bare white background. This is the strongest directional signal, backed by the most independent case studies. A clean kettle on white looks like fifty other kettles in the feed.
2. Real-feed-data motivators lift clicks when they are true and current. Price, % off, free shipping, and low-stock badges work when tied to actual feed values, and crucially they must update when the feed updates. This is exactly where a render-from-feed model beats a one-off Photoshop badge: a hand-made "−30%" chip goes stale and lies the day the sale ends; a feed-bound conditional badge disappears on the next refresh.
3. Legibility and restraint win. High contrast, large labels, one focal point. Overloaded frames suppress CTR by burying the product.
4. Branding helps, but the magnitude is brand-specific. Frame it as "test it," never as a guaranteed lift. For a deeper look at the template patterns themselves, see AI-designed catalog images.

Building it from your XML feed

If you have decided you want a designed image per product, the workflow on top of an imported feed is short, and it leans entirely on the "image is just a URL" fact above:

1. Import your existing product feed URL. Emberfeed parses the XML (standard RSS with the g: namespace) and stores each product.
2. Design one template: HTML and CSS (with a starting layout drafted by Claude or ChatGPT from the brief the editor composes), or a visual layer editor. Bind price, sale price, and badge conditions to feed fields.
3. Take the served feed URL (/api/serve/<feedId>). It is the same XML structure as your source, but every image_link is rewritten to the render endpoint, with the ?v= cache-buster attached.
4. Paste that served URL into Meta Commerce Manager as the catalog data source. Meta fetches it, sees normal-looking image URLs, fetches each one, and caches the render. No special Meta setting is involved. If the ad also has dynamic overlays or an image template switched on, Meta applies them to your render, so check the ad's creative settings.
5. Watch the next 7 days. Compare CTR and CPC against your previous raw-photo baseline.

For the full Meta setup walkthrough, see the Meta catalog ads use-case page. The honest scope, one more time: this needs your source feed, it is not self-hosted, and Meta still owns product selection. What you gain is control of the lever that lives in your feed, the pixels at the image_link URL, applied automatically across every product in the catalog.