AI watermarks—digital tags that identify AI-generated content—are becoming more common, and they may affect how people perceive messages, especially in personal and professional settings. The European Union recently introduced rules requiring companies to embed these watermarks in AI-generated content in a machine-readable format. This gives users the ability to detect if a message was created or edited by an AI. Some major AI developers, like Google and Anthropic, have already implemented such systems, while OpenAI had developed one earlier but delayed its use due to concerns about user experience. Research shows that even when people can't tell if a message was written by AI, learning that it was AI-generated can change how they view the sender. People often interpret the effort and care behind a message as a sign that they are valued. If that effort is replaced by a machine, it can feel like the emotional connection is missing, even if the message's content is sincere. This perception is particularly relevant in the workplace, where AI use is more accepted in routine tasks but less so in personal communication, such as giving praise or feedback. The use of AI watermarks also ties into a growing culture of surveillance. In workplaces, employees may become more skeptical of messages that are flagged as AI-generated, especially when those messages are meant to convey personal care or sincerity. If checking for AI fingerprints becomes a regular habit, even casual communication can be affected by doubt and suspicion. This effect is not limited to the workplace—private messages, like an apology, may also be viewed as less genuine if they are identified as AI-generated. AI watermarks do not always provide a complete picture of how a message was created. They only indicate that AI was likely involved, without specifying the extent of its role. For example, a message that was mostly written by a person and only polished by AI is different from one that was entirely generated by AI. However, the watermark cannot distinguish between these scenarios. Additionally, the rules around watermarking are not applied uniformly—short messages and basic editing tools like grammar checks are often exempt. This means that a simple text message may not carry a watermark, while a longer, more detailed message might. Ultimately, while AI can assist in communication, human relationships are built on trust, care, and honesty. Watermarking may not enhance these qualities but could instead create suspicion or reduce the perceived sincerity of a message. Some experts suggest that instead of relying on automatic detection, people should have the option to voluntarily disclose when and how they use AI. This allows for more context and transparency in human interactions, where trust and intention matter as much as the words themselves.