This study examines herding behavior in the LQ45 index over the period 2014–2023 using non-linear CSAD and CSSD models, while separating bull and bear market phases and incorporating COVID-19 episodes under wide and narrow definitions. The study aims to determine whether herding in Indonesia’s most liquid blue-chip segment is systematic, episodic, or absent. Weekly data from 15 consistently listed LQ45 constituents were analyzed using regime-based nonlinear regressions, with market return as the explanatory variable and dispersion measures as the dependent variables. The results show no evidence of systematic herding in the full sample or across the two sub-periods. Herding appears only episodically in 2015 based on the CSAD model, but this signal is not confirmed by the CSSD model, indicating that the episode is indicator-dependent rather than structurally robust. In bull markets, the positive and significant nonlinear coefficient points to anti-herding, suggesting that investors respond in more differentiated ways when market conditions strengthen. No significant herding is found during the COVID-19 period. These findings imply that investor behavior in the LQ45 index is not dominated by a persistent follow-the-crowd tendency, but rather by fundamentally guided and regime-sensitive price formation. The study contributes to the literature by providing long-horizon evidence on herding in an emerging market, showing that the critical distinction is not herding versus non-herding, but systematic versus episodic herding. The findings also suggest that large and liquid equity segments may preserve price efficiency except under specific stress conditions.