Generative AI is developing at a fast rate, and is capable of producing realistic images. This creates growing difficulties for users to be able to distinguish authentic visual content from AI-generated or manipulated media. In digital news environments, this creates a particular challenge when it comes to transparency, credibility and trust. The Coalition for Content Provenance and Authenticity (C2PA) has introduced a technical standard and a small number of provenance label designs intended to communicate the origin and authenticity of digital media. However, there is still limited research regarding the effectiveness of different provenance label designs and how such information should be visually communicated to end users. This study therefore investigates C2PA provenance label designs, using three different variations. This study aims to gain insights on how these labels can be designed to support users in correctly interpreting image origin and evaluating what information users find most relevant in digital news contexts. The study used a qualitative-centered mixed methods approach, combining both eye tracking and semi-structured interviews. Three labels were designed and embedded in a simulated news environment. Label A acts as the baseline click-to-expand C2PA icon, Label B is an always-visible color-coded label using icons indicating whether the image was AI-generated or authentic, Label C is an always-visible layered information label, presenting summarized provenance details. 24 participants took part in the interview phase, while 23 completed the eye tracking phase. The findings from the interviews showed that always-visible provenance labels were generally perceived as easier to understand compared to the baseline C2PA icon. Label B was most frequently identified by participants as the easiest to understand due to its clear wording and immediate visual differentiation between AI-generated and human-created content. In contrast, Label C was often perceived as more transparent, but also more overwhelming due to the amount of visible technical information. The eye tracking data further indicated that Label C generated the highest level of visual attention among participants, suggesting that more detailed provenance labels attracted greater visual attention. Overall, participants preferred concise provenance summaries that communicated the most relevant information directly.