With regard to maintaining editable text (within a Smart Object) a Layers-based approach would offer advantages over an AI-based one (at current) but if the task is a »one-off« it may not be worth the trouble anyway.
Setting the Type Layer to Blend Mode »Multipy« and using a Blend if-setting on a screening copy on the background image would be one option.
To distort the text along the seams one couold us the Filter »Displace« with a custom Displacement Map – the Filter Liquify would also be an option but seems cumbersome in a case like this. (The Filters should be used as Smart Filters.)