The z-offset is needed because the G28 zeroing process and the G29 auto level process don’t measure the bed, they measure limit switches and corner washers. In an ideal world, the z-offset would be set so that when the firmware sets Z = 0 it puts the tip of the nozzle at the bed surface. The assumption here is that the bed is a perfect plane and the corner washers are perfectly identical.
Since there is no margin for error (i.e. Z = 0 can’t be below the surface of the bed), any tests for correct setting of the z-offset should use a positive Z value and then measure the gap between the nozzle and the bed.
Pick any positive value for which you have a tool to measure. The firmware default is G21, mm units, so setting Z = 1.0 and using a 1mm feeler gauge will work if you are spot on with your first z-offset value. You should measure at multiple points on the bed including the center.
Most feeler gauge sets don’t have a lot of gauges on both sides of the 1mm so maybe Z = 0.8 might give you a couple of steps to either side. The metric set I have is .05 - 1.0 mm in .05 mm increments.
The goal of this exercise is to get the z-offset value “in the ball park”. Your final “measurement” should be with filament laid down in the first layer of a print. Typical first layer heights with a .5mm nozzle might range from .18mm to .425mm. You should be able to find articles and videos on “the perfect first layer” on the internet as well as numerous test prints.
What I would do is a G28 and G29 and then visually examine the nozzle gap at X, Y, Z of 20, 20, 2; 20, 260, 2; 260, 260, 2; 260, 20, 2; and 140, 140, 2. If the gap appears to be the same at all spots (and appears to be approximately 2mm) then test those five points with Z = 1 using your 1mm feeler gauge.