The affinities of extinct organisms are often difficult to resolve using morphological data alone. Chemical analysis of carbonaceous specimens can complement traditional approaches, but the search for taxon-specific signals in ancient, thermally altered organic matter is challenging and controversial, partly because suitable positive controls are lacking. Here, we show that non-destructive Fourier Transform Infrared Spectroscopy (FTIR) resolves in-situ molecular fingerprints in the famous 407 Ma Rhynie chert fossil assemblage of Aberdeenshire, Scotland, an important early terrestrial Lagerstätte. Remarkably, unsupervised clustering methods (principal components analysis and K-mean) separate the fossil spectra naturally into eukaryotes and prokaryotes (cyanobacteria). Additional multivariate statistics and machine-learning approaches also differentiate prokaryotes from eukaryotes, and discriminate eukaryotic tissue types, despite the overwhelming influence of silica. We find that these methods can clarify the affinities of morphologically ambiguous taxa; in the Rhynie chert for example, we show that the problematic “nematophytes” have a plant-like composition. Overall, we demonstrate that the famously exquisite preservation of cells, tissues and organisms in the Rhynie chert accompanies similarly impressive preservation of molecular information. These results provide a compelling positive control that validates the use of infrared spectroscopy to investigate the affinity of organic fossils in chert.