There's no bright line rule, in part because the concepts of "species" and "genus" are blurry and based on limited information. Even genetic analysis is an imperfect science that rarely compares whole genomes, and most species labels predate that kind of analysis.
Avocado is only graft compatible with about half the species in its genus, Persea, for example. But within the rose family, Rosaceae, it's very common for things in different genera to be graft compatible. Loquat (Eriobotrya japonica) and quince (Cydonia oblonga) are graft compatible, for example, as are things like rowan (Sorbus aucuparia) and European pear (Pyrus communis).
The only way to know for sure whether two species are compatible is to try, especially if you can't find any reports of them being incompatible. Many things have never been tested because other things already work well enough that no one has bothered trying more distant relatives.