Long-term P fertilization increases soil phosphorus (P) labile fractions, which can be associated with crop P uptake and grain yield and are useful to improve fertilizer recommendations. Research aims were to evaluate in long-term experiments with different P fertilization in a Mollisol and a Vertisol: i) the changes of soil P fractions, and ii) the relationship between soil P fractions with long-term P fertilization, accumulated apparent P budget, grain P, total P uptake, soybean (Glycine max L.Merr.) and maize (Zea mays L.) grain yield. Soil P fractions were measured after 1 and 9 yr since the beginning of long-term experiments. Experiments included an initial fertilization rate of 200 kg P ha -1 , and annual P fertilization rate of 36 kg P ha -1 . Bray1-P, total, organic, and inorganic P in fine (<53µm) and coarse (>53µm) (CF) soil fractions, and in NaHCO 3 extract were measured.Initial P fertilization increased inorganic and total P fractions. Whereas, Bray1-P, total P in NaHCO 3 extract and in the CF were the fractions with most increased with continuous long-term P fertilization in both sites. In the Mollisol, maize grain yield was unrelated to long-term P fertilization. In the Vertisol, total P in NaHCO 3 extract, and total and organic P in the CF were more closely related to soybean grain yield than Bray1-P. We proposed soil P indices of labile inorganic and organic P that showed close relationships with soybean grain yield and may be useful to improve the diagnosis of P soil fertility.