transforming data in a generalized linear model

So I am using a generalized linear model to analyze the results of animal location distances to a polygon created in ArcGIS. I have a bunch of distances, including zeros, when the animal is inside the polygon. The data is highly right skewed so I was using an inverse gaussian with a log link function but because it includes zeros I get an error message. If I log transform the data, it is normally distributed and I can run the generalized linear model with an identity link function and get results. My question is, is it valid to transform data for a generalized linear model? I'd thought most of the point of using these models was to avoid transforming data, but that may only be part of it. Doing a simple ANOVA doesn't work as there are too few degrees of freedom.