Phase Without Record: Closing a One-Sided Readout Loophole in Tests of Quantum-Mediated Gravity

Can a classical channel reproduce the phase signature normally interpreted as evidence for a quantum mediator? We show that it can—if only one conditional direction is measured. This work identifies a one-sided readout loophole in phase-based tests of quantum-mediated interactions and closes it using two direction-resolved Bargmann loops, (\mathcal Z) and (\mathcal Z'). Their product obeys the exact identity (\Delta_{\square}=\mathcal Z(\mathcal Z')^{*}), leading to an additive “record ledger” that reveals where which-path information is deposited. A classical feed-forward channel can hide its record from a one-sided audit, but not from the two-sided construction. We derive a tight separable ceiling that remains unchanged under product-form population imbalance and extend the certification to arbitrary unequal edge coherences through an exact two-qubit PPT semidefinite program. Readout-calibration uncertainty is incorporated through simultaneous Clopper–Pearson confidence regions and a conservative linear-program certificate. For experimental reproducibility, we provide a parameter-complete trapped-ion geometric-phase rehearsal, including pulse parameters, measurement settings, calibration procedures, closure scans, classical controls, and predeclared acceptance criteria. The conclusion is deliberately conditional: the protocol certifies a nonclassical locally mediating channel when the local-mediation premise is independently justified. It does not, by itself, constitute an experimental observation of quantum gravity or exclude every direct or nonlocal matter interaction.