We show that a large-area imaging survey using narrow-band filters could detect quasars in sufficiently high number densities, and with more than sufficient accuracy in their photometric redshifts, to turn them into suitable tracers of large-scale structure. If a narrow-band optical survey can detect objects as faint as i= 23, it could reach volumetric number densities as high as 10−4 h3 Mpc−3 (comoving) at z∼ 1.5. Such a catalogue would lead to precision measurements of the power spectrum up to z∼ 3–4. We also show that it is possible to employ quasars to measure baryon acoustic oscillations at high redshifts, where the uncertainties from redshift distortions and non-linearities are much smaller than at z≲ 1. As a concrete example we study the future impact of the Javalambre Physics of the Accelerating Universe Astrophysical Survey (J-PAS), which is a narrow-band imaging survey in the optical over 1/5 of the unobscured sky with 42 filters of ∼100-Å full width at half-maximum. We show that J-PAS will be able to take advantage of the broad emission lines of quasars to deliver excellent photometric redshifts, σz≃ 0.002 (1 +z), for millions of objects.