This paper presents an epistemological perspective of how web mining can be performed by human agents and a technical overview of a possible approach to achieve this. The concept of visual mining of information is based on the principle of creating a fused information space that could be navigated using visual perception, just as we do within a natural environment. It is argued that as the information available through human-created artifacts increase, conventional methods of information acquisition will fail to circumvent the bottleneck created by the associated information overflow. Four types of humancreated artifacts are identified: cognitive externalisations, artistic expressions, communicative accounts and factual records. It is the communicative artifacts that are responsible for the information overflow. As a possible way forward, it is suggested that an information space could be constructed in two layers, a perceptual layer and a cognitive layer. The information in the perceptual layer could be encoded in iconic cues to create an information landscape that could be navigated visually. Hence, allowing the sophisticated engine of visual perception to take charge of information navigation. The cognitive layer, on the other hand, will accommodate the information engendering data structures of the communicative artifacts, which will be structured in such a way that allows an efficient form of information consumption.