Digital signatures and message authentication codes are well known methods for ensuring message integrity. However, they rely on computations which are too hard to be performed by humans and are instead done on computers. Trusting a digital signature implies trusting the computer which produced/checked it. Often, this trust cannot be taken for granted. This paper presents a method for visual authentication of large messages which relies on embedding a human-recognizable watermark and needs practically no computational power on the receiver side. Also, using a simple challenge-response mechanism is proposed to prevent attackers from obtaining signatures without author's knowledge.