Context
Beyond the Vacuum formalizes generative engine optimization as a competitor-aware strategy-selection problem and evaluates combinations of rewriting strategies rather than optimizing a target document in isolation.
What changed
The work introduces a two-phase pipeline using Bayesian Optimization of Combinatorial Structures to search rewriting-strategy combinations, then derives preference pairs and grounded reasoning traces to train a model to propose combinations for a document corpus.
Why it matters
It provides evidence that GEO interventions can interact with the surrounding competitive corpus, so gains measured for one page in isolation should not automatically be treated as stable when other documents also optimize.
Visibility OS interpretation
Visibility OS interprets the study as support for competitor-aware experimental design in GEO. It does not establish a universal production ranking factor or behavior of any named commercial AI-search platform.
What remains uncertain
How well do competitor-aware strategies transfer to live production retrieval systems? How do effects change when multiple independent publishers adapt simultaneously? How durable are gains as corpora and engines change?