The ATP-dependent tubulin–tyrosine ligase (TTL) restores the carboxy-terminal tyrosine of α tubulin in αβ tubulin that has been previously detyrosinated. Here we show that the carboxy-terminal tetradecapeptide of detyrosinated α tubulin is used by TTL as a substrate, albeit at 50-fold lower efficiency than αβ tubulin. The minimal system provided by the TTL/peptide combination mirrors the TTL/tubulin system in all aspects tested, and shows a pronounced substrate inhibition. Synthetic peptides varying in length and/or containing single amino acid replacements were used to analyze the TTL specificity for the carboxy-terminal sequence of detyrosinated α tubulin. Peptides ending like α tubulin with the sequence Gly-Glu-Glu are optimally tyrosinated once a peptide length of 12 residues is reached. Position -1 of this recognition sequence, to which the tyrosine is added, must be glutamic acid. Position -2 accepts only an acidic amino acid but glutamic acid is by far preferred over aspartic acid. These results explain why a subpopulation of brain α tubulin, which ends with the sequence Gly-Glu, is not tyrosinated by TTL. The carboxy-terminal dodecapeptide of brain α tubulin with its polyglutamyl side-chain on position -6 shows the same substrate activity as the corresponding synthetic peptide lacking the side-chain. We discuss the substrate specificity of TTL for different α tubulins and speculate why tubulin is a better substrate than the optimal peptide covering the carboxy-terminal of detyrosinated α tubulin.