In annotation walking, analysis of a partially described biological system produces new information that helps to drive further advances in incremental fashion.

In systems such as natural products biosynthesis, CRISPR-associated (Cas) proteins, type 2 toxin-antitoxin systems, and capsular polysaccharide biosynthesis, the underlying bioinformatics grammar allows for varying selections from a guild of protein families to fill particular roles. As new systems are examined, new protein families will be detected that meet all validation requirements for guild membership, join annotation-driving databases such as TIGRFAMs, and define new molecular markers that can extend systems discovery even further.

In the ideal case, annotation walking leads to greater confidence in annotation toward the end of the bioinformatics journey than during the middle stages, as is the familiar experience of correctly solving jigsaw or crossword puzzles. In CRISPR/Cas systems, for example, most gaps between the originally defined markers cas1-4 have now been filled in by less universal members of the large guild of CRISPR-associated protein families, and all members of the guild would perform well in cross-validation experiments.

The term annotation walking is meant to evoke older terms "chromosome walking" and "primer walking."