Ancient DNA authentication is crucial for archaeological and evolutionary studies, but current methods face limitations including high cost, contamination risks, and computational complexity. We present a machine learning approach for human ancient mitochondrial DNA authentication using only FASTA sequences, bypassing the need for read-level data while supporting flexible age thresholds. Our method classifies samples as ancient/modern based on sequence features (CG-content, relative size, N-content, and normalized relative compression), achieving higher than 90% accuracy and F1-scores. This demonstrates that FASTA-based features alone can effectively distinguish ancient DNA, providing a scalable, non-destructive alternative to traditional techniques like radiocarbon dating or damage pattern analysis. The open-source tool is available at https://github.com/viromelab/amtDNA-Authenticator .