The strongest evidence to support the classical plume hypothesis comes from seismic imaging of the mantle beneath hot spots. However, imaging results are often ambiguous and it is questionable whether narrow plume tails can be detected by present-day seismological techniques. Here we carry out synthetic tomography experiments based on spectral element method simulations of seismic waves with period T > 10 s propagating through geodynamically derived plume structures. We vary the source-receiver geometry in order to explore the conditions under which lower mantle plume tails may be detected seismically. We determine that wide-aperture (4,000–6,000 km) networks with dense station coverage (<100–200 km station spacing) are necessary to image narrow (<500 km wide) thermal plume tails. We find that if uncertainties on traveltime measurements exceed delay times imparted by plume tails (typically <1 s), the plume tails are concealed in seismic images. Vertically propagating SKS waves enhance plume tail recovery but lack vertical resolution in regions that are not independently constrained by direct S paths. We demonstrate how vertical smearing of an upper mantle low-velocity anomaly can appear as a plume originating in the deep mantle. Our results are useful for interpreting previous plume imaging experiments and guide the design of future experiments.