Mass spectral libraries provide a tool for identifying unknown compounds using both molecular weight and fragmentation information. Mass spectrometers with electrospray ionisation (ESI) and atmospheric chemical ionisation (ApCI) sources have the capability to produce data of this type using in-source collisionally induced dissociation (CID), and in-source CID libraries can be created. Due to the variation in electrospray source design from different instrument manufacturers, the production of reproducible in-source CID spectra that can be used in libraries for all instrument types is not a trivial task. To date, the evaluation of the production of in-source CID libraries has tended to focus on similar instruments from one manufacturer. The studies have also tended to focus on specific compound classes, with a limited molecular weight range.This report describes the findings from the investigation of protocols for the creation of mass spectral libraries using ESI in-source CID on six instruments from four different manufacturers. The overall goal was to create a spectral library for the identification of unknowns. The library could then be applied across all manufacturers' electrospray instruments. Two different experimental protocols were attempted. The first used a tuning compound to establish standard ESI source conditions, with fixed fragmentation potentials. The second involved the attenuation of the [M + H](+) ion to a known degree. A diverse range of compounds (pharmaceutical, photographic, pesticides) was tested to establish the reproducibility of the spectra on the six instruments. Both protocols produced spectra on the various instruments that in many cases were very similar. In other examples, the spectra differed not only in their relative ion abundances, but also in terms of the spectral content. Important observations regarding the effect of ion source design are also reported. The degree of spectral reproducibility was calculated off-line by comparing the five most abundant ions (20% for each ion that matches) from each spectrum on each instrument. This approach was adopted, as we do not possess a software package that met our requirements for spectral comparison. Match factors (% fit) were calculated by comparing each spectrum against the spectra recorded for the same compound and then for all other compounds, on each instrument. The % fit values derived by the off-line approach gave a clear view of the spectral reproducibility from instrument to instrument and also discriminated the spectra of the various compounds from each other. The applicability of this approach was tested using a blind trial in which several compounds were presented as unknowns, their in-source CID spectra recorded and the five-ion approach used for identification.