Do banks extending government-guaranteed loans simultaneously reduce their risk exposure to firms? Using unique euro-area credit register data and the COVID-19 guarantee programs as a laboratory, we find that 1 euro of guaranteed lending was associated with a reduction of 28 cents in non-guaranteed credit, relative to other banks lending to the same firm. Substitution was highest for riskier and smaller firms in more affected sectors and for stronger banks. Nevertheless, banks offered cheaper credit and longer maturities to guaranteed loan recipients, especially more fragile ones. This improvement in lending terms is the flipside of credit substitution.