Recent correlational evidence implicates gratitude in personal and relational growth, for both members of ongoing relationships. From these observations, it would be tempting to prescribe interpersonal gratitude exercises to improve relationships. In this experiment, couples were randomly assigned to express gratitude over a month, or to a relationally-active control condition. Results showed modest effects of condition on personal and relational well-being. However, those whose partners were perceived as being particularly responsive when expressing gratitude at the initial lab session showed greater well-being across a range of outcomes, whereas this was not so for people in the control condition. Notably, evidence raises concerns about the effectiveness of artificial injections of gratitude when the partner is perceived to be low in responsiveness. Given the importance of close relationships, this work highlights the need for more theory-driven basic research tested in context before assuming what appears to work naturally will also work artificially.