The aim of the project is to identify wide common proper motion companions to a sample of spectroscopically confirmed M and L metal-poor dwarfs (also known as subdwarfs) to investigate the impact of metallicity on the binary fraction of low-mass metal-poor binaries and to improve the determination of their metallicity from the higher-mass binary. We made use of Virtual Observatory tools and large-scale public surveys to look in Gaia for common proper motion companions to a well-defined sample of ultracool subdwarfs with spectral types later than M5 and metallicities below or equal to ~0.5dex. We collected low-resolution optical spectroscopy for our best system, which is a binary composed of one sdM1.5 subdwarf and one sdM5.5 subdwarf located at ~1360au, and for another two likely systems separated by more than 115000au. We confirm one wide companion to an M subdwarf, and infer a multiplicity for M subdwarfs (sdMs) of 1.0_-1.0_^+2.0^% for projected physical separations of up to 743000au. We also find four M-L systems, three of which are new detections. No colder companion was identified in any of the 219 M and L subdwarfs of the sample, mainly because of limitations on the detection of faint sources with Gaia. We infer a frequency of wide systems for sdM5-9.5 of 0.60_-0.60_^+1.17^% for projected physical separations larger than 1360au (up to 142400au). This study shows a multiplicity rate of 1.0_-1.0_^+2.0^% in sdMs, and 1.9_-1.9_^+3.7^% in extreme M subdwarfs. We did not find any companion for the ultra M subdwarfs of our sample, establishing an upper limit of 5.3% on binarity for these objects.