The need for improved internet searching has grown in direct proportion to the exponential growth of the web. Thus, Google was born to dominate the world of search engines to this day with its deceptively simple searches. The scientific literature is growing apace and the days where a researcher can stay on top of their field by reading across the relevant journals are numbered. We increasingly rely on the highly filtered searches of one of the most efficient engines on the block: PubMed (www.ncbi.nlm.nih.gov/entrez/query.fcgi). The main drawback is that it covers only abstracts and keywords. To address this serious shortcoming, Nature Publishing Group revamped its own search engines (www.nature.com/dynasearch/app/dynasearch.taf) to provide full access beyond the firewalls and introduced semantic text matching (the section labelled 'more articles like this' on abstract pages). Furthermore, 650 publishers assembled to form CrossRef (www.crossref.org), a non-profit service with the important aim of adding permanence to the web-based literature through a system of digital object identifiers (DOIs; www.doi.org). CrossRef also launched a trial search engine (www.crossref.org/crossrefsearch.html) last year that permits full content searching of the participating 29 publishers (including NPG) on a Google platform.

In the meantime, Google ran with this idea, launching Google Scholar in November (still in its beta version at http://scholar.google.com). This system promises the Google magic for academic texts. Importantly, the algorithm ranks hits by webpage interconnectivity and weighs them further through inbuilt citation analysis, which is also displayed as a number against a hit. However, although Google citations follow similar trends to the citation standardbearer Thomson ISI, the numbers certainly do not match. Part of the discrepancy may be because Google Scholar has yet to index a significant fraction of the literature. It remains to be seen if the new kid on the block will become an authoritative port of call for coveted bibliometric data. A limitation at present remains a lack of search filters, such as date and article type, and a lack of a solid name/first name link, which will irk all Dr Smiths. With a nod to open-access publishing, Google Scholar throws open links to other versions of a given paper alongside links to the journal, PubMed and PubMedCentral. This feature has some overtones of the challenges to the music industry seen in online music-sharing services such as Napster a few years ago, and it remains to be seen if this will force publishers to rethink copyright. Meanwhile, Elsevier has launched its own search powerhouse, Scopus (www.scopus.com; see Nature 428, 683; 2004) which promises highly structured deep paper-trawling with citation statistics at a price.

Finally, we still want to throw in a word for good-old-fashioned journal browsing: bad search engines force you to browse, but through junk; good search engines circumvent all browsing by leading only to the expected. Nothing beats reading Nature Cell Biology cover to cover.