Background: Stroke is an increasing health problem world-wide. Atrial fibrillation (AF) is a major risk factor for stroke, and the attention towards AF screening is rising, while new monitoring-technologies emerge. We aimed to evaluate the performance of a large panel of screening strategies and to assess population characteristics associated with diagnostic yield. Methods: Persons with stroke risk factors but without AF were recruited from the general population to undergo screening with implantable loop recorder (ILR). New-onset AF lasting ≥6 min was adjudicated by senior cardiologists. After continuous monitoring for >3 years complete day-to-day heart rhythm datasets were reconstructed for every participant, including exact time of onset and termination of all AF episodes. Random sampling was applied to assess the sensitivity and negative predictive value (NPV) of screening with various simulated screening strategies compared to ILR. The yield across strategies and population subgroups was compared using non-parametric tests. Results: The rhythm datasets comprised 590 participants enduring a total of 659,758 days of continuous monitoring and 20,110 AF episodes. In this data, a single 10-sec ECG yielded a sensitivity (and NPV) of 1.5% (66%) for AF detection, increasing to 8.3% (67%) for twice-daily 30-sec ECGs during 14 days, and to 11% (68%), 13% (68%), 15% (69%), 21% (70%), and 34% (74%) for a single 24-h, 48-h, 72-h, 7-day, or 30-day continuous monitoring, respectively. AF detection further improved when subsequent screenings were performed, or when the same monitoring-duration was spread over several periods as compared to a single period (e.g. three 24-h monitorings vs. one 72-h monitoring), p<0.0001 for all comparisons. The sensitivity was consistently higher among participants with age ≥75 years, male sex, CHADS2 score >2, or brain natriuretic peptide (NT-proBNP) ≥40 pmol/L, and among participants with underlying ≥24-h AF episodes compared to shorter AF, p<0.0001 for all screening strategies. Conclusions: In screening for AF among participants with stroke risk factors, the diagnostic yield increased with duration, dispersion and number of screenings, although all strategies had low yield compared to ILR. The sensitivity was higher among participants who were older, males, or had higher NT-proBNP. Clinical Trial Registration: URL: https://clinicaltrials.gov Unique Identifier: NCT02036450.