Social media increase the transmission of false news, hoaxes and opinions, misleading, manipulating and misinforming the population due to the speed at which information spreads. Currently there are different websites known as fact-checkers that help to unmask news as Maldita or Newtral in Spain. They analyse certain news and publications in different media like Twitter, and report on their attempts to manipulate the population with accurate data. To do this, people in charge, analyse the news and look for reliable sources that help to contrast them, but as it is a tedious and manual job, it is not possible to deal with the large amount of information published on networks. The main objective of the work consists of analysing and understanding the emotional information of miss-informed texts in order to identify patterns and characteristics that could possibly be repeated in publications that seek to manipulate and provide false information. In this case we used tweets focused on the last political elections in different regions in Spain: Madrid, Catalunya, Castilla y León and Andalucía. The extracted characteristics, such as emotions, sentiments, toxicity and fakeness, have been obtained using pre-trained models. Finally, based on the characterisation of the texts and the veracity of each one, we train a model capable of approximating whether or not a tweet is a hoax based on the way it is written.