Tuberculosis (TB) is an infectious disease caused by Mycobacterium tuberculosis, which infects more than 23% of the world’s population. With the emergence of drug-resistant TB (DR-TB) and latent TB infection (LTBI), rapid diagnosis of DR-TB and LTBI has become a challenge for the prevention and control of TB. Herein, we highlight these challenges and discuss emerging clustered regularly interspaced short palindromic repeats (CRISPR)-based diagnostics in TB detection. Currently, the clinical diagnosis of M. tuberculosis infection mainly depends on pathogenic and molecular biological methods, including sputum smear, sputum culture, and Xpert. Although CRISPR-based diagnostics have not been applied to the clinical diagnosis of TB, they have shown exciting preponderances in TB diagnosis compared with traditional methods, including higher sensitivity, less sample input, and shorter turnaround time. CRISPR-based diagnostics represent a potential tool to address the challenges and natural weaknesses associated with traditional TB diagnosis methods. Based on the currently available data, we suggest that future CRISPR-based TB diagnostics should be developed in the direction of automation, modularization, diversification, and intelligence. By combining the CRISPR platform with various systems, such as microfluidic chips, droplet microfluidics, electrochemical techniques, and optical systems, the specificity and sensitivity of TB diagnosis may be revolutionized.