We propose a model of why implausible numeric suggestions function as anchors and influence peoples estimates. In two experiments we tested the hypothesis that peoples confidence in their knowledge modulates their susceptibility to implausible suggestions. We found that higher confidence reduces the impact of implausible anchors. A third experiment showed that knowledge, defined as participants agreement on a value, weakly affects the impact of a suggestion. These results indicate that theoretical accounts for the impact of implausible suggestions on estimations should consider the role of meta-cognitive information.