Built-in generators versus an independent one
Ad platforms increasingly generate assets themselves, from backgrounds and text variations to short videos built from stills. These features cost nothing extra, are tuned to the platform's own placements and are quick to use. They also work one platform at a time, keep their output inside that ad account, can change brand assets after anyone has reviewed them and leave little record of what was shown. An independent generator produces for every platform from one set of brand rules and approved claims, sends ads through approval before they reach any account, and keeps a record per asset. A common arrangement uses both: independent production for the core creative, and platform features only where their changes have been tested and allowed.
Platform-ready bundles, not loose images
Each ad platform wants a bundle rather than a picture: images in several ratios, headlines and descriptions within its character limits, logos, video in the lengths and orientations it serves, and sometimes a product feed. A useful generator outputs one bundle per platform, checked against the current specs before anyone uploads it, with ad names and tracking parameters that let analytics tie results back to the concept rather than to a file. For dynamic creative it delivers pools of components instead, each approved on its own, with rules for which may appear together. Either way, the output should arrive ready to traffic, not as a folder someone has to sort.
Policy, disclosure and claims checks before upload
Disapprovals cost learning time and, repeated, can put an ad account under restrictions, so check before upload. Generated ads have habits that breach policy or law: lines that assert or imply the viewer's personal attributes, such as health or financial difficulty, which Meta's advertising standards prohibit; false urgency, such as 'last chance' on an offer that never ends; and before-and-after images, which several platforms restrict. Disclosure is a separate decision per ad. Realistic synthetic people or events can fall under the EU AI Act's deep-fake duty, applying since 2 August 2026, platforms add their own AI labels, and political ads carry stricter synthetic-media rules. Record each decision with the ad.
Refreshing creative before it tires
Tired creative usually shows up in cost before anyone notices it in the ads themselves: the same audience and budget buying fewer results week by week. Keep a bench ready for that moment, meaning approved variants of the concepts that are working, so rotation happens before performance slides rather than after a scramble. Refresh in two tiers. New executions of a proven concept come first, because they reuse approved claims and pass review quickly; new concepts follow on a slower cycle with full review. Synthetic White applies approved claims and banned words at generation and keeps a per-asset record of the model behind each ad, so replacements start inside the same rules as the originals.
Updated 25 September 2026