You may even arrange it uniformly on the top width of the head, provided that the shearing force at the interface between web and flange protrusions stays in permissible range. It is easy to calculate, (assuming a negative moment bending situation) count the numbers of bars in the protrusion to one side, multiply for area and standing stress in the bars; do that in 2 sections 1 ft apart, near support; the difference between the two is the shearing force in such foot of an arm of concrete head; for the stress, divide by the area in such ft of the interface, i.e. 1 ft x thickness at root of arm, and compare the standing shearing stress with the factored or nonfactored stress permitted for the case. If too much, you can't put that number of bars in the protrusion; or more exactly said, you must reinforce in shear the shear interface to meet the code provisions; a shear-friction calculation may be permissible for the case but since the section will be for the factored loads likely cracked in tension make sure whatever the code says about compressive struts or limitations on the allowable shear force due to extenuation of the compressive capacity of the concrete for the situation is met. It is better to assume a reasonably low permissible stress in shear at the interface, say, 1.5 MPa under factored forces, or what the code allows, and if not met, place more main rebar in the web projection, and less outside.
All what above mainly meaning that the amounts can be quite precisely calculated; you only need a section analyzer that gives you the status of the section, in your case being of interest the stresses in the steel. I did some worksheets on Mathcad that give the status of T-beam sections under positive and negative bending action, and should still be available at mathsoft's site in the Collaboratory section for Mathcad 2000.
