Metrology and control of critical dimensions (CDs) are key to the success of nanotechnology. Modern nanotechnology and nanometrology are largely based on knowledge developed during the last 10 to 20 years of semiconductor manufacturing. Semiconductor CD metrology entered the nanotechnology age in the late 1990s. Work on 130-nm- and 90-nm-node technologies led to the conclusion that precision alone is an insufficient metric for the quality assessment of metrology. Other components of measurement uncertainty (MU) must also be considered: 1. sample-to-sample measurement bias variation, 2. sampling uncertainty, and 3. sample variation induced by the probe-sample interaction. The first one (sample-dependent systematic error) is common for indirect and model-based CD metrologies such as top-down and cross-sectional scanning electron microscopy (SEM) and scatterometry (OCD). Unless special measures are taken, bias variation of CDSEM and OCD could exceed several nanometers. Variation of bias and therefore MU can be assessed only if reference metrology (RM) is employed. The choice of RM tools is very limited. The CD atomic force microscope (AFM) is one of a few available RM tools. The CDAFM provides subnanometer MU for a number of nanometrology applications. Significant challenges of CDAFM remain, such as the following: 1. the finite dimensions of the probe are limiting characterization of narrow high-aspect spaces, 2. the flexibility of the probe complicates positioning control, 3. the probe apex sharpness limits 3D AFM resolution, 4. the lifetime of atomically sharp probes is too short, and 5. adsorbates change properties and dimensions of nanometer-sized objects considerably. We believe that solutions for the problems exist; therefore, we will discuss the role of RM in nanometrology, current RM choices, and the challenges of CDAFM as well as suggest some potential solutions.