Glossary
What are dynamic ads?
Dynamic ads automatically customise their creative elements in real time based on user data, behaviour or context — assembling the specific ad shown rather than serving a fixed asset.
How they work
A dynamic ad is a template plus a data source. The template defines the layout and the fixed elements; the feed supplies the variable ones — a product, a price, an image, a location, a message.
At serve time the system picks which values to use from what it knows about the context, and renders the combination. One creative definition can therefore produce thousands of distinct executions without anyone designing them individually.
Where they are used
- Product retargeting. Showing the specific items a user viewed, the most common application by volume.
- Catalogue advertising. Rendering from a live product feed so price and availability are current.
- Localisation. Store, language and currency varied by location.
- Dynamic creative optimisation. Testing element combinations at scale and shifting delivery toward the ones that work.
What makes them fail
Almost always the feed rather than the creative. A dynamic ad is only as accurate as the data behind it, and the failure modes are all visible to the customer: advertising a sold-out product, showing a superseded price, promoting something the user already bought.
The second failure is combinatorial. A template with five variable elements generates combinations nobody has reviewed, and some of them will be nonsense or worse — a cheerful message against an inappropriate product, a discount claim that no longer applies. Constraint rules matter more than creative variety.
Both are governance problems rather than design problems, and both get worse as the feed grows.
Dynamic creative on generative surfaces
Dynamic ads are the closest display precedent for how creative works inside an AI assistant, and the step beyond them is significant.
A dynamic ad assembles from a fixed template with fixed slots. A generative surface has no template — the assistant renders in its own voice and structure, so what an advertiser supplies is not a filled template but the structured facts the model will express. The layout, the phrasing and the emphasis are all decided at generation time.
That makes feed accuracy more important than it has ever been, because errors are no longer confined to a rectangle. A wrong price in a feed becomes a wrong price stated by the assistant, which is where hallucination liability starts.
Common questions
Do dynamic ads outperform static ones?
Usually in retargeting and catalogue advertising, where relevance to a specific product is the whole value. For brand messaging the advantage is far less consistent, and a single strong execution often beats many mediocre variants.
What do I need to run dynamic ads?
An accurate, frequently-updated product feed and a set of constraint rules. The feed is the hard part and the part that determines whether the output is any good.
More in ad formats and creative
The shapes an ad takes, and where it sits.