I don’t think there’s a programmatic way though blueprints to get what you want to achieve natievly. You might have to edit a bit of the engine to functionally “Add” ground truths to your tests.
It may be easier to have test geneate the screenshots, gather all the updated screenshots that failed through a script outside of blueprints and have someone review and update the tests.
Avoid tests that have dynamic elements and geneate scenes/levels specic for test cases.