Biomarkers are of interest for use in research and for the identification of the most valuable voice-related parameters in the clinic. The identified voice parameters eventually usable as biomarkers should be measurable and serve as indicators of normal biological activities, disease-related processes, or responses to therapeutic treatments. The clinical use of voice-related parameters was focused on in this study. Therefore, 2 literature searches were performed by The Royal Society of Medicine Library (UK) for the period of 2013–2023. Since many biomarkers are based on artificial intelligence (AI), the first search was made on voice and artificial intelligence. A total of 332 articles were found. The second one was carried out on the biggest group in the results: voice-related parameters Parkinson’s disease, but the analysis can be used for many voice-related disorders. Between 2013 and 2019, 47 papers measured voice-related parameters in Parkinson’s disease, of which 4 included some machine learning (ML). Between 2019 and 2023, 51 papers included voice-related parameters in Parkinson’s disease, of which 20 included ML, which is a huge difference from the previous group of papers. The first 47 papers found in the search on voice and Parkinson’s disease, mostly non-AI papers, were based on 7561 patients and 1513 controls from 2013 to 2019 (and 5 reviews on non-AI). The most used voice parameters in non-ML papers related to Parkinson’s disease, with the number of papers in parentheses, were fundamental frequency and standard deviation (40); jitter, absolute, and percent (29); shimmer, absolute, and percent (23); harmonics-to-noise ratio (23); voice handicap index (25); and intensity measurement (24). Other parameters are mentioned up to 14 times: Signal-to-Noise Ratio, Maximum Phonation Time, Spectrography, Cepstrum Analysis, Voice Range Profile, and the GRBAS test. Reviews conducted between 2018 and 2021 on voice-related parameters in Parkinson’s disease showed great heterogeneity between studies. Out of 98, 24 papers included AI, with 6488 patients and 531 controls, using 2–453 features to identify voice-related parameters. A total of 2 reviews on data sets, recording protocols, and signal analysis from 2022 and 2023 showed the issues with limited, unbalanced, and large differences between data sets. Well-defined tests, features, and data sets are necessary in the future to measure quantitative deviations of voice. Non-ML studies show clear differences in the measured parameters compared to healthy controls and also for treatment effect, but mostly the studies were not comparable, and the results were not quantitatively compared to other disorders. The artificial intelligence studies had too large a variety, especially of features and analysis of features.