This review identifies 426 sites with potentially suitable characteristics for tidal stream energy development, across nineteen countries in Europe, the Americas, Asia and Australasia. The most common site assessment quantifies the theoretical resource, which is the maximum amount of total energy that can be extracted by filling a site with turbines. This includes not only the energy extracted directly for electricity production, but also the energy losses resulting from support structure drag, drive train losses and wake mixing. The aggregated theoretical resource estimate is 1,000 TWh/year, from 262 sites (62% of those identified), across six countries. A more informative, albeit less common, indicator of electricity generation potential, is provided by practical resource studies, that estimate annual electricity production once economic, environmental, regulatory and social constraints are taken into account. Practical resource assessments are limited mainly to UK sites, with an estimated resource of 34 TWh/year, equivalent to 10% of the theoretical resource estimate. Of the seven countries with sufficient resource information, sites in the UK, Indonesia and New Zealand show the greatest potential to make national-scale electricity supply contributions. Resource assessment in France, Canada, USA and China indicates regional-scale impact potential, with possible further resource that may be developed, depending on future resource assessment of an additional 80 sites. For countries limited to resource assessment of ambient flow characteristics only, sites in Norway, Faroe Islands, Japan, South Korea and the Philippines show the greatest energy potential. This review brings to light the inconsistent and wide ranging site selection criteria and practical constraints that are adopted in the literature. This is reflective of uncertainty in (i) what constitutes a suitable site, (ii) how site suitability evolves over time (e.g., with changing competing energy costs in the energy sector), and (iii) the level of energy that can practically be extracted. When these uncertainties are combined, along with uncertainty in resource data, and the magnitude of energy losses, reported P10 and P90 resource estimates can lie 43% below and 30% above the P50 value respectively. It is recommended that future resource assessment characterises how the resource magnitude is impacted by site selection criteria/practical constraint ranges, that acknowledge and reflect their time-dependency and uncertainty.