We use a custom crawler to index scientific articles all around the web, just like Google Scholar. The coverage varies across fields. Preprint sources like arXiv are covered completely.
Depending on the data availability (open access), the assistant reads within metadata such as title and abstracts. Whenever possible, the fulltext is considered in the search.
No, Poia Science does not distribute any articles. It solely provides the metadata so that an article can be easily located with an institutional access provider.
No, the virtual assistant literally reads through "big data". On one hand, the data wouldn't even fit on your laptop, on the other, the search wouldn't finish within any reasonable timeframe. This task is best left to our beefy cluster.