AbstrakIndustri pinjaman online mulai berkembang di Indonesia pada tahun 2016. Terdapat dua jenis pinjaman online yang berkembang di Indonesia yaitu pinjaman online ilegal dan pinjaman online legal. Bertambahnya jumlah kasus pinjaman online ilegal berdampak pada menurunnya tingkat customer trust masyarakat Indonesia terhadap industri pinjaman online. Tujuan penelitian yaitu melakukan analisis semantik untuk melihat behaviour pihak perusahaan pinjaman online dari isi pesan pinjaman online yang terdapat dalam UGC (User Generated Content) yang menyebabkan banyak konsumen pinjaman online mengalami penurunan tingkat kepercayaan kepada perusahaan pinjaman online. Penelitian ini menggunakan text mining yaitu analisis semantik. Analisis semantik akan dilakukan dengan menggunakan software Wmatrix5. Data diperoleh dari hasil crawling menggunakan Google Collab dan web scraping Phantombuster pada sosial media Instagram dan Twitter.Hasil analisis menunjukkan terdapat 15 kelompok semantik yang ada dalam pesan pinjaman online, kelompok tersebut antara lain yaitu crime (G2.1-), giving (A9+), paper documents and writing (Q1.2), knowledge (X2.2), polite (S1.2.4+), knowledgeable (X2.2+), unmatched (Z99), law and order (G2.1), getting and possession (A9+), money: debts (I1.2), personal relationship: general (S3.1), speed: fast (N3.8+), helping (S8+), information technology and computing (Y2), dan business: selling (I2.2). Kata Kunci: Semantik Analisis, Pinjaman Online, Kepercayaan Pelanggan, Perilaku Pelanggan, Wmatrix5. AbstractPeer to peer lending industry began to develop in Indonesia in 2016. There are two types of peer to peer lending industry that are developing in Indonesia, namely illegal and legal. The increasing number of cases of illegal peer to peer lending industry has an impact on the decline in the level of customer trust of customer peer to peer lending in Indonesia.The purpose of the research is to conduct a semantic analysis to see the behavior of peer to peer lending companies from the content of peer to peer lending messages contained in UGC (User Generated Content) which causes many peer to peer lending consumers to experience a decrease in the level of trust in peer to peer lending companies. This research uses text mining, namely semantic analysis. Semantic analysis will be carried out using Wmatrix5 software. The data is obtained from crawling using Google Collab and web scraping Phantombuster on Instagram and Twitter social media.The results of the analysis show that there are 15 semantic groups in peer to peer lending messages, these groups include crime (G2.1-), giving (A9+), paper documents and writing (Q1.2), knowledge (X2.2), polite (S1.2.4+), knowledgeable (X2.2+), unmatched (Z99), law and order (G2.1), getting and possession (A9+), money: debts (I1.2), personal relationship: general (S3.1), speed: fast (N3.8+), helping (S8+), information technology and computing (Y2), and business: selling (I2.2). Keywords: Semantic Analysis, Peer to Peer Lending, Customer Trust, Customer Behaviour, Wmatrix5.