Generative artificial intelligence (GenAI) is increasingly used in school informatics, yet technically correct assistance can still interfere with the reasoning through which learners develop algorithmic competence. This conceptual paper focuses on upper-secondary algorithmisation, defined here as the staged design and representation of algorithms from problem interpretation and decomposition through pseudocode or flowcharts to debugging and verification. It proposes pedagogical noise as an integrative, taskspecific diagnostic lens for identifying misalignment between an AI contribution and the competence-forming work that should remain learner-owned at a particular stage. A structured and traceable conceptual synthesis compares this lens with cognitive offloading, productive struggle, scaffolding failure, automation bias, performance-learning dissociation, feedback overload, reduced epistemic agency, overreliance, hallucination, and academic misuse. The synthesis develops seven proposed forms of pedagogical noise, each tied to a primary diagnostic dimension: completeness, timing, transparency, learner judgement, causal debugging, volume/actionability, and ownership/accountability. Worked cases show how technically correct GenAI support may function as scaffolding or substitution depending on learner state, timing, output granularity, transparency, and agency. The contribution is not a new theory of learning or a claim that GenAI is inherently harmful; it is a bounded diagnostic framework and a set of design propositions for preserving learnerowned reasoning in upper-secondary algorithmisation.