SUMMARY Archaeomagnetism provides important constraints to help us understand the past behaviour of the geomagnetic field. For archaeologists, archaeomagnetic dating has become a potential supplement to traditional dating methods (e.g. radiocarbon dating). Although China has a long history with numerous archaeological discoveries, the collection of archaeomagnetic data remains scarce. In this paper, we provide new archaeomagnetic directions from four late Neolithic (c. 2000 BC) sites in Shandong province, China. After a careful characterization of magnetic mineralogy and a detailed alternating-field demagnetization of the oriented samples, a total of nine archaeomagnetic directions (each with both declinations and inclinations) were obtained, which fill the large gap at c. 2000 BC in the Chinese palaeomagnetic secular variation (PSV) curve. Combining these new results with previously published data, we updated the Chinese PSV curve for the last 7 kyr. We have compared the updated curve with several global geomagnetic models (e.g. pfm9k.1a, ARCH10k.1, CALS10k.1b). Comparisons show that the CALS10k.1b model does not yield a reasonable fit of the data, and the fit becomes worse for older intervals. This poor fit could be explained by the fact that the CALS10k.1b model consists of a large amount of sedimentary data. Therefore, the PSV pattern is difficult to determine due to the strong aliasing effect. On the contrary, the ARCH10k.1 model gives a much better fit than other models because its data are mostly from archaeological materials and the data are mainly from the Northern Hemisphere. The field intensity and PSV are potentially correlated, with a weak field corresponding to an enhanced PSV. However, due to the lack of data for certain time intervals, the proposed correlations need to be further tested. To explore if PSV exhibits longitudinal symmetric or latitudinal antisymmetric patterns like those of the geocentric-axial-dipole (GAD) model, we compiled and compared data from three East Asia countries (China, Korea and Japan) and from four areas (East Asia, North America, Europe and the Middle East, and Australia and New Zealand) of which the latitudes are between 30° and 40°. In the East Asia region, the PSV patterns shown in each data set are consistent because of the geographic proximity of these three countries. However, when comparing the PSV curves from the four global areas, we suggest a potential declinational minimum between 0 AD and 2000 AD. Although further confirmation and investigation are needed, this declination minimum could be diachronous, sweeping from East Asia to Australia and New Zealand, and then North America. Future studies should focus on adding more reliable and precisely dated data to better delineate the PSV trends. Archaeomagnetic dating is promising when a PSV curve can be continuously reconstructed.