If AI systems are to be used effectively humans need to trust them. In part this is due to whether they are sufficently accurate or reliable, but even the best AI system will not be used unless people also believe it to be reliable. Transparency and explainable AI can help to generate trust, just as we are more trusting of people who do not seem to be hiding things. In addition, the way in which AI is embedded within user interactions can change the extent to which it is trusted, and also the robustness of the complete human–AI system, for example by presenting multiple reuslts with confidence values rather than a single answer.