How an image asset works
You upload a picture once and it is stored in the account’s asset library, where any campaign that needs imagery can draw on it. Google then crops and places it to fit whatever slot is being filled, which might be a wide strip on a news site, a square in a feed, or a tall block on a phone.
Because the same picture has to serve those different shapes, the platform asks for more than one crop — commonly a square, a landscape and a portrait version — and will crop automatically if you do not supply them. It also allows several images per campaign, and chooses between them at serving time much as it chooses between headlines.
The same word appears in a different context on the search side, where an image asset attaches a small picture beside a text ad. That is a separate placement with its own rules, and it should not be confused with the imagery driving a responsive display ad.
Why image assets matter
On Display, Demand Gen and Performance Max, the picture is doing the work that a keyword does in search. Nobody asked to see your ad, so the image is what decides whether the message is read at all, and a weak one loses the impression before a single word is considered.
It also carries most of the trust. A photograph of the real premises, the real team or the actual product tells a stranger something a stock photo cannot, and that difference matters more, not less, in a market where buyers are cautious about who they pay online.
Common mistakes with image assets
Overlaying text on the picture is the mistake that keeps recurring. The system crops to fit, so words placed near an edge are cut off, and a half-visible slogan makes an otherwise decent ad look broken. Put the message in the text fields, which are never sliced.
Uploading a single wide photo and letting the platform find a square inside it comes next. Automatic cropping does not understand composition, and it will happily remove the face, the product or the sign the picture was chosen for.
Generic stock imagery is the quiet failure. It rarely offends anybody and it rarely persuades anybody either, and when several competitors licence the same library the ads become interchangeable. Low-resolution uploads and heavily filtered images fail for a related reason: they read as amateur at exactly the moment a stranger is deciding whether you are real.
How to act on it
Shoot or select with the crops in mind, keeping the subject near the centre so every version survives. Supply the square, landscape and portrait cuts yourself rather than delegating that decision. Give each campaign several genuinely different pictures — a product, a person, a place, a result — instead of near-identical frames from the same session, because sameness gives the system nothing to choose between.
Then read the reporting. Each image is rated against the others in the same campaign, so you can see which one earns its place and replace the weakest with a different idea rather than a similar shot. Refresh before the images tire rather than after, and keep the library tidy so nobody rebuilds a campaign around a photo of last year’s premises.