Describe the issue linked to the documentation

fit_intercept in Ridge says "(i.e. data is expected to be centered)", but the "data" is ambiguous. From some experimentation, it seems that the "data" is the y in the regression problem. This is confusing because most resources say penalized models should not be fit with an intercept, so users may be tempted to pass fit_intercept=False, producing undesirable results.

Suggest a potential alternative/fix

Explicitly reference what data is assumed to be centered when fit_intercept=False and explain that when fit_intercept=True that it is not used in the optimization, in accordance with resources on the topic.

I can handle this if I have interpreted what fit_intercept=False does.