In this study we use sedimentary ancient DNA metabarcoding from two marine sediment cores. The first Kastenlot core MSM05/5-712-2 was taken at western Fram Strait (subarctic North Atlantic) from which we collected 12 samples including one biological replicate. The second Kastenlot core SO201-2-12KL was retrieved from Kamchatka Strait (subarctic North Pacific) from which we collected 9 samples and 2 samples were collected from a pilot core taken next to the Kastenlot core. Total DNA was extracted from approximately 2ml sediment per sample in 3 batches with up to 9 samples and one extraction blank. Total DNA was concentrated and if necessary diluted to 2.5 ng/µl. For each batch we performed PCRs in triplicates including a PCR no template control (NTC). We amplified a diatom-specific, 76 bp long part of the rbcL gene with tagged primers Diat_rbcL_705F (AACAGGTGAAGTTAAAGGTTCATAYTT) and Diat_rbcL_808R (TGTAACCCATAACTAAATCGATCAT). The PCR-products were purified and pooled in equal concentrations. The sequencing library was prepared with the Mid Output kit v. 2 according to the Fasteris Metafast protocol for low complexity amplicon sequencing and checked by qPCR. The library was sequenced (2 x 150 bp, paired-end) on the Illumina NextSeq 500 at the Fasteris SA sequencing service (Switzerland). For two samples we sequenced 4 PCR-products and for two samples we could only get 2 PCR-products with sufficient DNA content for sequencing. Here, we provide the reads with our tagged primers removed for all sequenced PCR-products including extraction blanks and NTCs. The sample names are composed as follows: CoreName_Depth_PCRproductID. The PCRproductID contains the experiment number (e.g. IE031P) so that negative controls can be associated with the samples of the corresponding PCR run.