A Semantic Search Engine could be described as an agent that allows users to search at the topic level. Once a topic of interest is identified, then drill down to the specific document. Semantics is been applied these days by search engines in a few closely related fields. A possible way, is by using statistical analysis to retrieve semantic matches to a query. With this approach, they are capable of providing relational results based on the topics and keywords given by the user, and help them in the proccess of narrowing the search to be more specific.

To outline the process, there are basically 3 steps to consider: (1) User enters a search query, with a given number of words describing wath the topic of their search is, (2) The engine returns results matching the words user entered, plus adding suggested alternatives (3) The second step repeats itself untill the results returned are sasticfactory to the user.

Automated semantic tagging is also possible, when the topic of a given page is well known. In this application, dynamic categorization of topical information about a query is implemented without any human intervention. These categorizations could take place in real time or be maintained incrementally over a time period. This is possible because there are available cognitive objects and conceptual models to identify sense information. Then these could be precisely weighted and measured, to determine the relevance among variable words forms and grammars in multiple corpus of content.

Google's application of semantics goes beyond Adsense. Relational Search is now evident when returning results from their index. This takes place if you get search results alternatives based upon an initial query you performed or after pulling pre categorised phrases on a database. Not so generalized yet due to multiple current limitations like speed, scalability and others. Performance is getting better and better, but it will be a matter of years before the big search engines start annotating webpages purely by using semantic ontologies.

There is a very distinctive focus in the current search engine algos towards AI. Google is the most obvious in the application of such technologies, but is not the only one. Some of the things search engines are utilizing in their machine learning process includes studying user behaviour, recording actions, drawing conclusions.

The more traditional ways that search engines have used previously to establish the populatiy of a given site, have been among others: volumen of inbound links, and their quality score based on relevancy of their anchor text and co-relation to the theme of the target sites. There are several factors involved in evaluating the links data that helps search engines to determine the popularity of a site and the importance of its content. In this new approach, the engines have been able to gain a finite amount of aggregate usage data from its various programs and services that allows them to further develop algos into more accurate and relevant services.

Nowadays, geotargeting or the personalisation of search results based on the geographic location of the users, browser-based and social bookmarking are just a few new elements added to the equation. Traffic source information to given search results also count. This allows searh engines to really evaluate sites based on how popular they truly are, and weight their content and position in SERPs.