Doesn't have to, I think. Text is actually quite adequate. Sure, images are great but that's only for a subset of alerts, like AMBER, but that can be achieved by secondary distribution channels like a dedicated webpage or social media. There won't be network congestion during those alerts. You don't really need images or video for alerts like floods, hurricanes, or even missile incoming alerts. For those you only need an address or coordinates that can point people to the closest shelter. Yeah, not all people are familiar with the surroundings (like traveling) but in that kind of scenarios every people around you will pool together and get to the same location.

The whole core functionality that the alert system must achieve is a near 100% delivery rate that uses minimal bandwidth of the network. You don't want to stress the network because there are more important traffic need to be routed, like 911.