An original biosensor was developed for the direct conductometric detection of ochratoxin A (OTA) in olive oil samples. The biosensor is based on thermolysin (TLN) immobilization into a polyvinyl alcohol (PVA)/polyethylenimine (PEI) matrix containing gold nanoparticles (AuNPs) and cross-linked at the surface of gold interdigitated microelectrodes using glutaraldehyde. Under optimal conditions (35 min cross-linking time, working pH of 7 and temperature of 25 degrees C), the biosensor response was linear up to 60 nM OTA, with a sensitivity of 597 mu S mu M-1 and a limit of detection of 1 nM. This value was 700 times lower than the detection limit obtained using the more classical method based on enzyme cross-linking in the presence of bovine serum albumin (BSA). PVA/PEI hydrogel creates a very favorable aqueous environment for the enzyme. In addition, interactions between protonated amino groups of PEI and negative charges of both citrated AuNPs and thermolysin improve their dispersion in the polymer blend, favoring enzyme stabilization and accessibility of the substrate. No conductometric signal was observed after OTA injection in the absence of AuNPs, in agreement with the insulating properties of the cross-linked PVA/PEI hydrogel film. Incorporation of AuNPs into the TLN/BSA biomembrane helped improving the sensitivity by 5.3 but this latter remained 140 times lower than the sensitivity of TLN/AuNPs/(PVA/PEI) biosensor. The study of enzyme kinetics showed that V-i vs [OTA] plot exhibited a non-hyperbolic trend, indicating that kinetics does not display a Michaelis-Menten behavior. Biosensor response times are longer (7-48 min) comparatively to TLN/BSA biosensor with a maximal value of 25 min. This difference is due to the diffusion phenomenon through the pores of the polymer membrane. The proposed OTA biosensor was very reproducible with a relative standard deviations (RSDs) in the 3-15% range and stable over a 30-days period when stored at 4 degrees C in 20 mM phosphate buffer between two measurements. The method was further evaluated using commercial doped olive oil samples. No pretreatment of the sample was needed for testing and no matrix effect was observed. Recovery values were close to 100% demonstrating the suitability of the proposed method for OTA screening in olive oil.