One big challenge with active targeting of nanocarriers is non-specific binding between targeting molecules and non-target moieties expressed on non-cancerous cells, which leads to non-specific uptake of nanocarriers by non-cancerous cells. Here, we propose a novel Trojan-horse targeting strategy to hide or expose the targeting molecules of nanocarriers on-demand. The non-specific uptake by non-cancerous cells can be reduced because the targeting molecules are hidden in hydrophilic polymers. The nanocarriers are still actively targetable to cancer cells because the targeting molecules can be exposed on-demand at tumor regions. Typically, Fe3O4 nanocrystals (FN) as magnetic resonance imaging (MRI) contrast agents were encapsulated into albumin nanoparticles (AN), and then folic acid (FA) and pH-sensitive polymers (PP) were grafted onto the surface of AN-FN to construct PP-FA-AN-FN nanoparticles. Fourier transform infrared spectroscopy (FT-IR), dynamic light scattering (DLS), transmission electron microscope (TEM) and gel permeation chromatography (GPC) results confirm successful construction of PP-FA-AN-FN. According to difference of nanoparticle-cellular uptake between pH 7.4 and 5.5, the weight ratio of conjugated PP to nanoparticle FA-AN-FN (i.e. graft density) and the molecular weight of PP (i.e. graft length) are optimized to be 1.32 and 5.7 kDa, respectively. In vitro studies confirm that the PP can hide ligand FA to prevent it from binding to cells with FRα at pH 7.4 and shrink to expose FA at pH 5.5. In vivo studies demonstrate that our Trojan-horse targeting strategy can reduce the non-specific uptake of the PP-FA-AN-FN by non-cancerous cells. Therefore, our PP-FA-AN-FN might be used as an accurately targeted MRI contrast agent.