Abstract We provide an economic assessment of zero-rating offers in the context of mobile internet access services and draw six lessons: (1) Zero-rating can have several different characteristics that crucially affect their economic and welfare assessment. Thus, regulatory interventions must be based on a careful case-by-case analysis. (2) In the context of zero-rating offers, it is often crucial to evaluate the extent to which users are able to activate and deactivate a (throttled) zero-rated tariff option. If activation/deactivation is easy and instantaneous, a sound economic theory of harm for consumers will in many cases be hard to establish. (3) Similarly, if access to zero-rated partner programs is non-discriminatory and entails low barriers to entry, a sound theory of harm for content providers will usually not be given. (4) Zero-rating can be beneficial for consumers and (legal) content providers alike by contributing to a reduction of illegal content. Combined with throttling it can mitigate congestion problems. However, by requiring all content belonging to the same content category to be treated equally with respect to throttling, independent of whether a content provider opted for zero-rating or not, the existing regulation creates a negative externality on those content providers that do not wish to be zero-rated for some reason. (5) Particular attention should be paid to the impact of throttled zero-rating tariffs on the competition between mobile network operators (MNOs) and MVNOs. The latter may not be able to compete on equal footing with MNOs, because they benefit less from the traffic management aspects of zero-rating. (6) Competition among (infrastructure-based) ISPs provides a safeguard against severe rent extraction and, thus, an abuse of throttling and zero-rating as an exploitative device. Therefore, regulators should carefully account for the competitive environment and the existing tariff portfolio and options before deciding to intervene. Competition policy, rather than ex-ante regulation, may be more suitable for this task.