Corpus Engineering
Shaping the whole body of evidence an AI system can retrieve about a brand, rather than optimising individual pages. Attributed to Cody C. Jensen and Searchbloom, 2026.
What it actually means
A model does not read your page. It reads a corpus, of which your page is one document among many, most of which you do not control. Corpus engineering treats that whole body as the unit of work.
In practice that means auditing what exists about you across the web, finding the contradictions and gaps, and setting out to change the aggregate rather than perfect one document.
Where it shades into manipulation
The honest version of this is closing gaps in a record that under-represents the truth: getting accurate information into places that carry stale or missing facts. The dishonest version is manufacturing corroboration.
The line is whether the evidence would exist if nobody were optimising. A genuine review, a real case study and an independent mention survive that test. A network of thin domains built to agree with you does not, which is why the same field also had to name Wasteful Domains.