Context. The Gaia DR3, released in June 2022, included low-resolution BP/RP (XP) spectra that have been exploited for the classification of various types of emission-line objects using machine-learning techniques. The Gaia Extended Stellar Parametrizer for Emission-Line Stars (ESP-ELS) algorithm identified 273 sources as potential planetary nebulae (PNe). Aims. We analyzed the sample produced by ESP-ELS to investigate the true nature of the objects classified as PNe. Methods. We extracted all sources from the Gaia catalog classified as PNe and filtered out 200 objects with secure classifications available in the literature. Of these, ∼65% correspond to known Galactic or Magellanic compact PNe, and ∼20% to D- or D'-type symbiotic systems. Via visual inspection of the XP spectra of the remaining sources, we identified a subset of 14 promising candidates with strong emission features attributable to Hα and [O III] λ5007. Although typical of PNe, such features are also consistent with Dor D'-type symbiotics, which are known to mimic compact PNe. We obtained spectroscopic follow-up observations for these objects with the 2SPOT facilities in Chile and France, and analyzed archival photometric data to further constrain their nature. Results. We report the identification of nine bona fide or likely D- or D'-type symbiotic systems, one PN in the LMC, one polar cataclysmic variable, and three possible Be stars in (or in the direction of) the SMC within our sample of 14 objects.