Abstract We present an analysis of the sensitivity of current and future LHC searches for new spin-0 particles in top–anti-top-quark $$ \left(t\overline{t}\right) $$ t t ¯ final states, focusing on generic axion-like particles (ALPs) that are coupled to top quarks and gluons. As a first step, we derive new limits on the effective ALP Lagrangian in terms of the Wilson coefficients c t and $$ {c}_{\overset{\sim }{G}} $$ c G ~ based on the results of the CMS search using 35.9 fb −1 of data, collected at $$ \sqrt{s} $$ s = 13 TeV. We then investigate how the production of an ALP with generic couplings to gluons and top quarks can be distinguished from the production of a pseudoscalar which couples to gluons exclusively via a top-quark loop. To this end, we make use of the invariant $$ t\overline{t} $$ t t ¯ mass distribution and angular correlations that are sensitive to the $$ t\overline{t} $$ t t ¯ spin correlation. Using a mass of 400 GeV as an example, we find that already the data collected during Run 2 and Run 3 of the LHC provides an interesting sensitivity to the underlying nature of a possible new particle. We also analyze the prospects for data anticipated to be collected during the high-luminosity phase of the LHC. Finally, we compare the limits obtained from the $$ t\overline{t} $$ t t ¯ searches to existing experimental bounds from LHC searches for narrow di-photon resonances, from measurements of the production of four top quarks, and from global analyses of ALP–SMEFT interference effects.