There have been research driven and market driven attempts at developing robust machine translation programs for most of the main languages in the world. The future development of reliable translation systems depends on the systematic evaluation of available programs showing the degree to which they can effectively transfer the message. This study aims at conducting a multi-level analytic evaluative examination of a number of available Arabic into English machine translation programs focusing on the functional characteristics and sub-characteristics of the output: reliability, fidelity, terminology and syntax. This evaluation proposes to develop metrics related to the four functional criteria. The data was selected from five text genres: technical, legal, literary, journalistic and economic which were translated from Arabic into English using available machine translation systems for Arabic. Google, Ajeeb, Professional Translator, 1-800-translate and Worldlingo are free, whereas Trasphere and An-Nakel are fee-based. The translated texts were assigned scores and then "diagnostically" analyzed trying to identify errors, deficiencies, limitations and loss of meaning. The overall evaluation was concerned with the current theme of accuracy trying to determine the fitness of a system within an operational context. This case study started in 2006 and ended in 2008. During this period the seven systems of translation were visited and revisited by the researchers to check and upgrade the output.