We show that there is "no stable free field of index $α\in (1,2)$", in the following sense. It was proved in a previous work by the authors, that subject to a \emph{fourth moment assumption}, any random generalised function on a domain $D$ of the plane, satisfying conformal invariance and a natural domain Markov property, must be a constant multiple of the Gaussian free field. In this article we show that the existence of $(1+\varepsilon)$-moments is sufficient for the same conclusion. A key idea is a new way of exploring the field, where (instead of looking at the more standard circle averages) we start from the boundary and discover averages of the field with respect to a certain "hitting density" of Itô excursions.