Evaluates retrieval and citation capabilities of generative engines on synthetic entities, indicating performance nuances.
Results record for the preregistered study deposited as DOI 10.5281/zenodo.21705818. The preregistered 2x2 (entity density × content structure) is declared non-interpretable due to a length-parity deviation, reported under the deviation clause. What survives, from confirmatory runs plus an independent verification subset, is that consumer ChatGPT and Gemini both perform live retrieval and citation of an orphan synthetic entity when handed its URL. A flagged instrument event is reported: on Gemini the synthetic-entity control does not isolate per-page retrieval, an index-level namespace association robust to logout, private mode, and IP change. Reported as a limitation and methods lead.
No takes yet. Share an insight, caveat, or question.
Ferreira et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: