Actually, I appreciate that feature very much. I like that when I search for Python, Google knows as a programmer I mean the language not the snake. Doubtless there are herpetologists who appreciate that Google knows they mean the snake.

Try and search anything even remotely technical or specific. Google brings back a few thousand irrelevant results, often with stuff like 'not found: word you were actually after' underneath in grey. Sometimes even when you do add quotes or speech marks around the term.

A better Google would be one that doesn't assume people want 'any old possible answer' and tries to answer specific queries as easily as general ones.

While there seems no barrier to entry, just consumer habit, google is faster than it used to be, despite a larger web and more users.

Speed of search results has been shown to be extremely important to users. Google's many large server farms, close to users, is a huge capital investment. Google has doubled down on this advantage, with google suggest, which requires even more power. For example, google has much lower latency than DDG. Replicating those server farms would be expensive; I'm not sure than even all of AWS could do it.

When I first used Google in 2000, I immediately knew they would come to dominate search, because their product was so much better. It's not enough to be as good as Google, you have to wipe the floor with them.

As others have pointed out, the Amazon Echo is a viable approach to doing this. Shame it sucks for anything but the most trivial queries right now.

Perhaps the answer is to do what Craigslist did. In a few categories, ones you have to ask for such as "Shopping" or "Jobs", there are ads, or product ordering boxes, clearly identified as such. Non-commercial information is not polluted with ads.