In general, users who access Web sites do not necessarily have the same abilities in terms of cognitive, sensory and physical conditions. An accessible Web site is one whose content can be used by as many users as possible; and to improve accessibility, the use of evaluation tools has been incorporated among other resources. However, from the point of view of the end user, the evaluation obtained by most of these tools does not fit to the individual needs and abilities of the users. On the other hand, there is a greater understanding of the need to automate the support to reach higher levels of accessibility. Promoting this type of support on both the developer and the end-user sides is a complex procedure that can be done by incorporating intelligent features, such as the collaborative participation of intelligent agents. In this paper, we introduce a multi-agent system based on visually impaired user’s profiles to improve Web Accessibility assessment. The proposal is validated through a proof of concept, by contrasting results against 10 wide-known assessment tools. Our results show that evaluations are more focused, and therefore, unnecessary assessments are avoided improving efficiency.