Last updated September 21, 2026. Here is how the engine works, what the guardrails are built to catch, where they fall short, and which parts stay your call as the advertiser.
Each drop is rendered by generative image and video models, conditioned on the brand brief you set during onboarding — your palette, your props, your storefront, your tone. Scenes are generated for your business rather than taken from a stock library or a fixed template with your logo dropped on top.
Generative models are probabilistic. Two runs on the same brief produce different images, and output can contain artefacts, odd details or text that is subtly wrong. That is the nature of the technology, not a fault we can fully engineer away.
No creative reaches your channels until someone on your side clicks approve in the control room. Adronis does not auto-publish creatives you have not seen. Approving is the moment the creative becomes your ad — so read it, including any price, offer, date or claim rendered into the image, before you approve.
Before a creative reaches your approval queue it passes automated checks aimed at the known failure modes of generative models. They are designed to filter out:
Creatives that fail a check are regenerated before the drop is finalised.
These checks are automated and statistical. They reduce the failure rate; they do not eliminate it. We do not guarantee that every creative reaching your queue is free of errors, invented detail or third-party material. Things that can get through include rendered text that reads correctly to a classifier but wrongly to a customer, an unintended resemblance to an existing image or mark, or a detail that is fine in general but wrong for your business.
That is why approval sits with you, and why our terms make clear that you remain the advertiser and the party legally responsible for what you publish.
Where a connected platform offers an AI-content label through its publishing interface and supports it for the format we are posting, Adronis sets that label automatically when it publishes on your behalf.
Two limits worth knowing. First, platform labelling is only as good as the platform's own interface — where a platform does not offer a label, or changes how it works, we cannot apply one. Second, disclosure rules differ by country, sector and ad type, and they change. We do not monitor which rules apply to your business, and applying a platform label does not by itself make an ad compliant. If you are unsure what disclosure your market or sector requires, take advice before you approve.
Reference photos and brand assets you upload condition the output. Only upload material you have the rights to, including permission from anyone who appears in a photo. The engine is not intended to generate images of real, identifiable people other than from material you have supplied and cleared.
We do not use your brand material to train generative models for other customers, and we do not sell it. What we keep, for how long, and who processes it is set out in the Privacy Policy.
Creatives, dashboards, metrics and figures shown across this site are illustrative examples of how the product works. They are not a forecast or a promise of results for your business.
Kill it in the queue rather than approving it, and flag it from the control room or through our contact page. If you believe a creative infringes someone's rights, write to ip@adronis.app and we will investigate.