Marketers’ use of AI influencers is growing. Although AI influencers on social media can achieve outcomes such as word of mouth (WOM), recent studies also suggest that AI influence may be in part due to novelty. Using a 2 (influencer type: AI vs. human) x 2 (message type: agitator vs. synthesiser) between-subjects experimental design, this study of 294 US Instagram users explores whether AI versus human influencers’ posts on Instagram, have different influence on customer engagement, WOM and source credibility. We also explore whether the communication style (agitating/synthesising) motivates these outcomes differently. Furthermore, as AI influencers may be eerie or unsettling for consumers, we consider whether self-congruence with the influencer (AI/human) mediates these relationships. We show that human influencers have a higher credibility and are perceived as more self-congruent with the actual-self. However, AI influencers are as effective as human influencers in driving outcomes such as engagement and WOM, and regardless of their post’s communication style. We also reveal a mediation effect of actual self-congruence between the type of influencer and these outcomes. Results provide support for the use of AI influencers, and we suggest ways for marketers to enhance the effectiveness of AI influencers, through appealing to actual self-congruence.