No, they don't but I know what you are referring to.

Most of the time I get the result:

  No results found for [phrase]
  Results for [phrase] (without quotes)

..but the thing is that I know websites containing [phrase] exist. Often many of them and they aren't "dark web" either. Google used to be able to find them but no longer is.

This gets more confusing because sometimes it does work. Namely if you look for phrases which are popular search queries e.g. if you search for..

  "hit me baby one more time"

..it will work amazingly well. But for other phrases it won't work at all and you will get the no results reply instead. It used to work for all phrases independent of their search popularity.

But yes, my original post was wrong. It does work for a lot of phrases.. but not for others.

I guess Google's engine dynamically optimizes things, and only indexes often searched phrases.

EDIT: Okay, this assumption was wrong too. I did some experiments to confirm my theory and the results showed that it is wrong..

It seems to work [0].

[0] https://www.google.com/search?source=hp&ei=f6xdWs2KJq7k_Qb0k...

This is what it looks like to me: https://imgur.com/zre6HdT

Probably this is because such pages do not exist, except for this page. Or at least DDG thinks so: https://duckduckgo.com/?q=%22It+never+rains+on+a+Wednesday+i...

Indeed a very nice way to prove this thread wrong :-)

I had also wondered whether its ability to match exact phrases has degraded. It would make sense, since keeping individual words on a distributed index is a lot easier than keeping a long phrase. But I had no way of telling, not having benchmarked it in the past.

Can they not index bigrams or trigrams, then chain together index hits? E.g. "It never rains in december" would hit on "It never rains", "never rains in", and "rains in december". Any result that hits on all of the indexes is not guaranteed to hit on the entire phrase but it would be a good candidate for the top result. The longer the phrase, the more likely a candidate hitting all necessary index phrases would match the exact phrase. This would at least put a limit on how large the indexes need to get.

Further, if they retain copies of the full text in their database they could do a filtered scan of the documents that hit on all subphrases to guarantee exact match. I could see that having too much of a performance impact at scale though.

In any case the dumbing down of Google search over the last few years is immensely frustrating to me.