Lifelogging augments people's ability to keep track of their daily activities and helps them create rich archives and foster memory. Information workers perform a lot of their key activities throughout the day on their desktop computers. We argue that activity summaries can be informed by eye-tracking data. Therefore we investigate 3 heuristics to create such summaries based on screenshots to help reconstruct people's work day: a fixed time interval, people's focus of attention as indicated by their eye gaze, and a reading detection algorithm. In a field study with 12 participants who logged their desktop activities for 3 consecutive days we evaluated the usefulness of screenshot summaries based on these heuristics. Our results show the utility of eye tracking data, and more specifically of using reading detection to determine key activities throughout the day to inform the creation of activity summaries that are more relevant and require less time to review.