The study considers translator training for handling so-called defective specialized texts that contain a range of errors. In Russia, it is conventional to use perfect tutorial texts written by native speakers and free of any grammar, lexical, or factual errors. However, in real life, a translator has to handle imperfect, defective texts. Nowadays, a translator of technical/research/other specialized texts is directly responsible for identifying and correcting source text errors. That is why this aspect shall be included in the translator training university-level programs. We propose a 5-tier hierarchical text verification algorithm called ‘the text verification pyramid’. Under the algorithm, both the source and target text are verified for compliance with the laws of nature, laws of logic, principles of operation applicable to the specific item considered, text cohesion requirements, and facts. The proposed verification approach is applied twice: first, to the source text, and then to the target text. The pyramid tiers are used from bottom to top, starting with the compliance with the laws of nature. The proposed text verification algorithm will improve translation quality. It is suggested to teach translators how to verify texts with the algorithm to promote critical thinking in terms of both the source text and the translation.