The total amount of polycyclic aromatic hydrocarbons (PAHs) in soils, given by exhaustive chemical extractions, does not relate directly to environmental risk, since only a fraction may be accessible to soil organisms. The rapid PAH desorbing fraction (Frap), which is weakly and reversibly sorbed to soils, is called the bioaccessible fraction, and can be estimated by non-exhaustive aqueous extractions. In order to better estimate Frap, different mild-extractants were tested, such as various cyclodextrins, surfactants and butanol. Their extractability performances were correlated to the Kd partition coefficients of seven PAHs obtained through sorption isotherms from five soils, but also to the PAHs molecular size and to the amounts of organic matter and of some clays (smectites and kaolinites). If hydroxypropyl-β-cyclodextrin was actually a good extractant to assess PAH accessibility, the polymer of carboxymethyl-β-cyclodextrin (pCMCD) was better (with a lower cost) to estimate the rapid mass transfer between soil particles and the soil solution, depending also on soil ageing. But Frap, estimated through pCMCD extractions, did not reflect the biodegradation of the PAHs after three months in soil microcosms. The chemical method underestimated the dissipation of 3–4 ring PAHs and overestimated that of 5–6 ring PAHs. So biodegradation was not only limited by PAHs mass-transfer, but also by biological factors, favoring the access of microorganisms to residual strongly sorbed fractions of 3–4 ring PAHs, and inhibiting the degradation of accessible but highly toxic 5–6 ring PAHs.