This paper argues that an adaptive extended Kalman filter (EKF) improves performance of Global Navigation Satellite System (GNSS) position, velocity, and timing (PVT) by single point positioning without sensor aiding or coupling in dense urban environments. One of the most important as well as difficult problems is the negative impact of non-line-of-sight (NLOS.) Large buildings shade line-of-sight (LOS) signals and create NLOS signals by reflection and/or diffraction so that the GNSS receiver eventually tracks the NLOS signal only, which causes significant and unexpected measurement errors that degrade the PVT performance of conventional positioning methods. To reduce this negative impact, an adaptive EKF is implemented for single point positioning. Six laps of test drives prove that the adaptive EKF drastically reduces position and velocity errors and removes outliers compared to conventional EKF. Regarding the horizontal position, for example, the adaptive EKF produces 16.1 m of cumulative probability in the one-sigma error range (68.27%), while the conventional EKF results in 48.0 m. Mean error is also minimized, from 12.90 m to 2.74 m. Similar improvements are present in vertical position and 3-D velocity. The only difference between the EKFs in this paper is the adaptive estimation of the covariance matrix of the measurement noise. Detailed analysis confirms that the adaptive covariance matrix of the measurement noise matched the actual measurement error, as all correlation coefficients exceed 0.95, which highlights significant improvement in positioning performance. The adaptive estimation of the covariance matrix has a simple formulation and the process is not expensive, which means it can run on a low-cost receiver in real-time. Thus, the adaptive EKF could be proposed as a simple and effective technique to reduce the negative impact of NLOS and to improve the GNSS PVT performance in heavy NLOS environments for any type of GNSS receiver.