Numerous archeomagnetic studies have provided high quality data for both the direction and the intensity of the geomagnetic field, essentially in Europe for the last 10 millennia. In particular, Greece supplies a lot of archeological materials due to its impressive cultural heritage and volcanic activity, so that numerous data have been obtained from burnt clays or historical lava flows. The most recent Greek secular variation curves are available for the last 8 millennia for the intensity and the last 6 millennia for the direction. Nevertheless, the coverage still presents several gaps for periods older than 2500 BC. In an effort to complete the Greek curve and extend it to older times, we present the archeointensity results from three Neolithic settlements in Northern Greece. The samples are of two different natures: burnt structures from Avgi (5250 ± 150 BC) and Vasili (4800 ± 200 BC), as well as ceramics from Dikili Tash (4830 ± 80 BC) and Vasili (4750 ± 250 BC). The samples have been subjected to standard rock magnetic analyses in order to estimate the thermal stability and the domain state of the magnetic carriers before archeointensity measurements. Surprisingly, very few ceramic samples provided reliable archeointensities whereas samples from burnt structures presented a very good success rate. Complementary studies showed that a detailed examination of the matrix color, following archeological information and classification standards can be a decisive test for pre-selection of sherds. In spite of these unsuccessful measurements from ceramics, we obtained an intensity value of 73.5 ± 1.1 μT for Dikili Tash, a higher value than the other data obtained in the same area, during the same period. However we do not have evidences for a technical artefact during the experiment. The burnt structures yielded two reliable archeointensities of 36.1 ± 1.8 μT and 46.6 ± 3.4 μT for Avgi and Vasili, respectively. Finally, we achieved a new archeomagnetic dating for these sites by comparing these new archeointensity values, combined to the directional measurements already published, with the Bulgarian secular variation curve. These new results contribute to extend the Greek secular variation reference curve towards older periods.