Koalas are a cryptic species that are surprisingly difficult to survey, especially in tall remote forests. We are undertaking a large-scale survey of koalas in forested environments of north-east New South Wales to assess their current status and response to timber harvesting. So far we have deployed SongMeters (SM4) at >100 sites over an extensive area, each for a period of 7 nights, to record male bellows during the breeding seasons of 2015/16. Surveys are targeting modelled moderate-high quality habitat in forests with a stratification based on time since logging and logging intensity. Recordings are scanned by Ecosounds software at QUT and then we manually checked all computer matches of koala bellows to rule out false positives. Our sample from a wide range of forests in 2015 found that koalas had a probability of detection of 0.3-0.4, indicating that low detection must be accounted for before relating occupancy to habitat type or forest successional stage.  The seven nights of acoustic data are highly suitable for accounting for imperfect detection prior to occupancy modelling. Preliminary results reveal naïve occupancy levels of 50-80 % across a broad range of forests; such high detection rates were unexpected based on results from previous survey methods. Once imperfect detection is accounted for, we will estimate probability of occupancy for a range of forest types in relation to timber harvesting. Acoustic detection is proving to be a highly successful and efficient technique for recording koalas in forested areas where traditional surveys have had limited effectiveness.