The influence of pre-Columbian human populations on Amazonian ecosystems is being actively debated. The long-standing view that Amazonia was only minimally impacted by human actions has been challenged, and a new paradigm of Amazonia as a "manufactured landscape" is emerging. If such disturbance was the norm until just 500 years ago, Amazonian ecosystems could be far more ecologically resilient to disturbance than previously supposed. Alternatively, if the "manufactured landscape" label is an overstatement, then policy that assumes such resilience may cause substantial and long-lasting ecological damage. We present paleoecological data suggesting a middle path, in which some areas were heavily modified, but most of Amazonia was minimally impacted. Bluffs adjacent to main river channels and highly seasonal areas appear to have been the most extensively settled locations. Away from areas where humans lived, their influence on ecosystems was very local. Consequently, we see no evidence suggesting that large areas at a distance from rivers or in the less seasonal parts of Amazonia were substantially altered by human activity. Extrapolating from sites of known human occupation to infer Amazon-wide landscape disturbance may therefore potentially lead to unrealistic projections of human impact and misguided policy.