Saliva is non-invasive, repeatable, and anatomically close to oral squamous cell carcinoma (OSCC), but proximity does not establish that a salivary biomolecule originates from the tumor. Cytokines, extracellular vesicles, coding and non-coding RNAs, and tumor-derived DNA may arise from tumor and immune cells, adjacent mucosa, periodontal tissues, salivary glands, oral microbiota, or systemic circulation. This review evaluates whether salivary immune signals reflect the OSCC tissue immune microenvironment using three evidence levels: quantitative paired tissue-saliva concordance within patients (Level I); indirect clinicopathological, source-enriched, paired-biofluid, or longitudinal support without quantitative paired tissue analysis (Level II); and saliva-only case-control differences (Level III). IL-6, IL-8/CXCL8, IL-1β, and TNF-α show relatively consistent salivary changes, whereas evidence for IL-10, selected chemokines, extracellular vesicle-associated PD-L1, and secreted miRNAs or lncRNAs remains limited or inconsistent. Salivary tumor DNA can show strong tissue-genotype concordance, but reflects tumor-derived material rather than immune state. Most immune candidates remain at Level II or III; none robustly meets Level I. Periodontitis, mucositis, infection, smoking, tumor ulceration, salivary flow, and sample processing complicate interpretation. We propose paired collection of tumor, adjacent mucosa, saliva, and plasma; spatial immunophenotyping; oral-inflammation controls; longitudinal perioperative and treatment sampling; standardized assays; and multicenter validation. The framework prioritizes biological traceability over diagnostic discrimination. At present, saliva is best considered a complementary longitudinal biosample, not a substitute for tissue immunophenotyping.