Background There are few data available regarding the use of on-demand pre-exposure prophylaxis (PrEP) for HIV prevention. We aimed to assess PrEP effectiveness, adherence, and safety in adults using daily or on-demand PrEP. Methods We conducted a prospective observational cohort study (ANRS PREVENIR) at 26 sites in the Paris region, France. We enrolled HIV-negative adults (aged ≥18 years) at high risk of HIV infection who were starting or continuing PrEP. PrEP was prescribed as a fixed-dose combination of tenofovir disoproxil and emtricitabine (245 mg and 200 mg, respectively, per pill). PrEP could be prescribed as a daily regimen with one pill per day or, in men who have sex with men (MSM) or in transgender women who have sex with men, as an on-demand regimen following the IPERGAY dosing recommendation. At enrolment and every 3 months thereafter, participants were tested for HIV and provided information regarding the PrEP dosing regimen used. Adherence to PrEP was assessed by self-report and by tenofovir diphosphate concentrations in dried blood spots. The primary outcome of HIV-1 incidence was assessed using Poisson regression among participants who started PrEP. This study is registered with ClinicalTrials.gov, NCT03113123, and EudraCT, 2016A0157744. Findings Between May 3, 2017, and May 2, 2019, 3082 people were assessed for eligibility and 3065 participants were enrolled. 3056 (99·7%) of 3065 participants reported using PrEP and were included in the analyses. The median age was 36 years (IQR 29–43), 1344 (44·0%) of 3056 participants were PrEP-naive, and 3016 (98·7%) were MSM. At enrolment, 1540 (50·5%) of 3049 participants opted for daily PrEP dosing and 1509 (49·5%) opted for on-demand PrEP dosing; these proportions remained stable during follow-up. Median follow-up was 22·1 months (IQR 15·9–29·7) and incidence of study discontinuation was 17·6 participants (95% CI 16·5–18·7) per 100 person-years. At the data cutoff on Sept 30, 2020, there had been six HIV-1 seroconversions (three participants using daily PrEP and three using on-demand PrEP; all were MSM) over 5623 person-years. Overall HIV-1 incidence was 1·1 cases (95% CI 0·4–2·3) per 1000 person-years, and did not differ between participants using daily PrEP and those using on-demand PrEP (incidence rate ratio 1·00, 95% CI 0·13–7·49; p=0·99). Four participants (two using daily PrEP and two using on-demand PrEP) discontinued PrEP due to treatment-related adverse events (nausea [n=2], vomiting and diarrhoea [n=1], and lumbar pain [n=1]). Interpretation In this study, which enrolled mainly MSM, HIV-1 incidence on PrEP was low and did not differ between participants using daily PrEP and those using on-demand PrEP. On-demand PrEP therefore represents a valid alternative to daily PrEP for MSM, providing greater choice in HIV prevention. Funding ANRS/Maladies Infectieuses Emergentes, Gilead Sciences, SIDACTION, and Région Ile de France. Translation For the French translation of the abstract see Supplementary Materials section.