In rare cases I want to search something on the Internet and I am pretty sure that the answer is there, but I am not able to formulate my search to get there (although I am an experienced searcher).

My last choice may be creating a new question where an experienced user tells me that this topic is treated in another question, with a question title that I would have never come across with. Or I get no answers.

Some things I noticed that complicate the search:

  • The keywords have multiple meanings (e.g. I want so know something very special about the "in" keyword in a programming language but almost every website contains an "in")
  • My answer gets overlaid by plenty similar but more general answers (because search results are listed by popularity I guess)
  • The tagging on the thread I search for is really bad

What are your tips?

  • 6
    Sometimes you have no other option then to click past the first 100 pages in a search result. Also on SO it is not bad if your question turns out to be a duplicate because if you included what you searched for, the next person that searches for that same phrase will find your question and then be guided to the actual answer. Signpost questions are important to have. – rene Oct 18 '20 at 12:41
  • 2
    "Google fu" is a skill, and needs a bit of experience and observation about how these search engines work, and how they can be triggered to come up with the wanted results. First of all: Ditch attempts to ask something in natural language, rather just use some essential keywords in the right order of relevance, and give them some specific context. For instance it's hard to find duplicate questions at Stack Exchange sites using the internal search engine, but google yields really good results at the point, if you prefix your search using site:<my preferred se site>.com. – πάντα ῥεῖ Oct 18 '20 at 13:10
  • 1
    ^ BTW, I am wondering why the SE devs didn't boost their search engine based on that yet. – πάντα ῥεῖ Oct 18 '20 at 13:11
  • 1
    You coulsd use the code search parameter to limit results to in when it is only used in code: stackoverflow.com/… – rene Oct 18 '20 at 14:06
  • 1
    Use correct terminology, add a category or use quotation marks that are certain to provide an example of what you are interested in, also check out the FAQs and related links (right column) to these Q&As: meta.stackexchange.com/a/66378/282094 meta.stackexchange.com/a/87900/282094 - minimum search is a few letters, you'll need some commonly accompanying text in addition to your extremely short term. - adding "keyword" and tagging with the programming language you are asking about is a good start. If you really try and don't succeed then others would fail too; so no dupe. – Rob Oct 18 '20 at 16:02

Ask on the site's per-site Meta, or, if your question is about the Stack Exchange network itself, right here. Tag your question and , if possible.

One of the major functions of Metas is to help users formulate questions and answers. Most people who participate in a Meta on a regular basis understand the subject matter of the main site (citation needed), possibly more than you do. What you are doing is, in effect, asking for help asking a question, possibly with a hope that you will be able to identify your potential question as a duplicate before you even post it. The Meta will help you with that.

For example, you might formulate a question for a site Meta as follows:

I need help with the following question, but I'm not really sure how to formulate it or if it is even on-topic. I suspect it might be a duplicate, but I haven't found anything. My question so far is "I'm trying to do the thing that is sort of like reticulating the spline, but with a parallelized matrix gradient where n>0. What are the steps to complete the process if my matrix is fully enveloped in threespace?" Can someone help me formulate this question and/or identify if it might already be a dupe?

A user might respond,

What you are asking about is called "Recogitation of the Matrix", and we use the tag here for that topic. We have a set of canonical questions because it is a frequently asked-about topic. You said that n>0, but we also need to know if it is <=1. See if any of the following questions fits yours:

How do I recogitate the matrix when 0<n<=1?

How do I recogitate the matrix when n>1?

Why is knowing the value of "n" so important when trying to recogitate a matrix?

I keep hearing that the second-order determinant of the recogitate of a matrix gradient is undefined if n=0.5, but everyone seems to treat it as a given and never provide a source. What's the proof?


Don't give up easily! Search and read the search results and search again and document your research, so that when that last choice of asking a new question comes you can demonstrate your research. In particular, if you are worried that your topic of interest is hidden by more popular but less specific sites, add more specific keywords (like programming languages, frameworks) or try to eliminate from the search results ("-" for google) the popular topics that you are not interested in.

Read about the topic in general/broaden your research. The more you know about the topic in general, the better you will be able to search. You will get more keywords right, you will give better context.

Ask people in chat (for the target stackexchange) if they can help you formulate useful search terms.

Take a step back. Maybe the question is so difficult to formulate because it's still a bit too unclear or too broad. It may need details or focus. Can you divide the question into smaller steps? If so, do that and search for sub-steps.

Finally, if everything else fails, ask that question, demonstrate your research and if it gets closed as duplicate - you get an answer and it may additionally serve as a search target for others. I wouldn't downvote a question that demonstrates that much effort, i.e. shows search effort and background knowledge, just for being a duplicate.

Personal anecdote: One or two questions I had fully formulated in the "ask question editor" and only then I got the keywords right, searched one last time, found a solution (on a stackexchange) and clicked cancel.

For completeness also a summary of search hints from the comments below this question by πάντα ῥεῖ, rene and Rob.

  • just use essential keywords in the right order of relevance in searches and give them specific context
  • prefix your search in google using site:<my preferred se site>.com
  • with the internal SE search engine use the code parameter to search for keywords only in code sections example
  • try to use correct terminology, use a category or quotation marks to specify an example that describes your situation
  • also read What is the XY problem? and How to ask a smart question

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .