He said: "The reason we feel pretty good in terms of competition is because what we're seeing and everything we're building today is built on top of the foundation of core web search rankings.
"If I say, 'Show me the Eiffel Tower', I want pictures of the Eiffel Tower but, if I say, 'Show me the money', I don't want pictures because I'm talking about the line from the Jerry Maguire movie. Google actually knows that because of web rankings telling us."
Google Glass, the company's wearable device which packs the capabilities of a smartphone into spectacles that allow hands-free operation, is among the first of a new generation of mobile technology it hopes will spur people to talk to its services. Google Now will also target smart watches, cars and the living room, allowing people to ask the television questions.
Apple has apparently recognised Google's advantage. This month it paid more than US$200 million ($245 million) to acquire Topsy, a start-up focused on finding patterns in the 500 million tweets posted on Twitter every day.
Observers, including Nick Halstead, chief executive of Topsy's British rival DataSift, have speculated that Apple will use the language analysis technology to improve Siri's understanding of queries before it processes them or delegates them to third-party web services.
Huffman said: "It does feel like a bit of a race. For us the race part often has a lot to do with [the engineering] talent, a little more than finding a start-up with a magic idea that we have to buy before anyone else."
But as well as recruiting the best software engineers, Google has been forced to implement a cultural shift towards accepting there are certain things about the real world that cannot be boiled down to an algorithm. It has assembled a team dedicated to cataloguing information and verifying information so Google Now can answer questions directly rather than provide a list of relevant web links as in a traditional web search.
Huffman said: "Where the humans come in is identifying which sources of data we should use. The other area is verifying data so that when people ask Google a question like, 'How old is Tom Cruise?' we've got the correct answer.
"We have ... accepted there are cases you need human validation if we're going to answer questions directly."