A system and related techniques detect the initiation of a user's search input and monitor that input character-by-character, to generate suggested search terms on the fly. Arbitration logic may monitor the user's keyboard of other entry of search or query terms at a Web search site or other search engine or resource, and examine that input along with stored query history or usage data on a real-time basis to predict or infer search terms which the user is attempting to transmit. Spelling corrections may likewise be made. According to embodiments, the arbitration logic and/or query history or usage data may be hosted in the user's machine, in the search service itself or in other resources. As the arbitration logic generates suggested search terms in inline fashion, those search suggestions may be presented to the user in real-time, for example by way of a wordwheel, drop-down or other dialog or interface. The user may for example choose to select one of the set of search suggestions before completing the typing or other inputting of their search data, and in embodiments may edit those suggestions in real-time as well. According to embodiments in a further regard, the user may provide configuration inputs to the automatic inline search suggestion feature, for example to turn that capability on or off on a per-session, per-search, permanent or other basis.