12

Given the drop in new question activity

A rising then declining graph showing new questions per month from 2008 until January 2026 SEDE source

and discussions like these:

I think it suffices to say that SO/SEs growth phase is past its prime. But as with any other website, it was expected to happen at some point.

Why a change of focus towards curation might be in order

There are only so many questions one can ask, even with new technologies/versions coming out. Even if there were no new questions at all, it wouldn't mean the "knowledge-bank" built here is complete:

  • Questions can be refined
  • Accepted answers can become outdated and need replacing/updating
  • Tags need (re-)organizing
  • Search (both the feature and how content is organized) need improvements

and much more (feel free to add to the list).

The company so far focused its developer manpower on optimizing new user experience/ new question creation. Perhaps effort could or should be diverted onto curation. After all, SE already has a community not unlike Wikipedias' "Wikipedians" that is willing to contribute and maintain the content here.

Unlike Wikipedia the bar to participating in curating content feels quite a bit more difficult (I admit that this is a subjective statement):

  • The most repeated issue I read was onboarding (for curation activities) is regarded as lacking.

  • Of note is also that necessary actions such as editing are locked behind sometimes high reputation point barriers, e.g. editing needs 2000 points. With the increasing content standard over the years and the drop of activity on the site, this makes it harder for new users to get points/privileges. Some users such as myself find themselves discouraged from curating.

There are more points that could be added (again feel free to add your own), yet so far all attempts that go towards improving curation went seemingly nowhere. Most notably:

In 2023 staff asked about the current reputation system, but despite 71 answers to that post, I cannot recall any significant change made to the reputation/privilege system since that time (as of writing that's almost 3 years), besides allowing 1 rep users to chat.

While I'm at it, here's a question for any staff that might stumble upon this: Is SE perhaps already planning a shift in its development priorities to move to curation?

The closest thing I got to finding something hinting to this shift was this comment on Upcoming initiatives on Stack Overflow and across the Stack Exchange network (October 2024):

As part of our user activation strategy, we have internally discussed a few different things related to improving the experience for reviewers and/or contributors. That being said, those are still internal discussions. When we are ready to focus on this and ask for feedback we’ll share more with the community. – Hoid

Seems to me like at some point something along these lines was at least being discussed?

Improve this question
3

3 Answers 3

Reset to default
9

It's a massive drop. Even if it were resticted to Stack Overflow (and it's not), then a ~95% drop in questions is going to affect things network-wide. At this point, I don't think there can be any alternatives: it's primarily due to people turning to AI instead of Stack Exchange. AI is still improving and becoming more widely used, so it may continue to drop. (Or maybe AI gets enshittified, and the opposite happens. I don't know; I can't predict the future.)

Actually, I noticed this on Wikipedia:

The convenient explanation is "AI summaries." I'm skeptical. What we're witnessing is something more profound: a generational shift in how people relate to knowledge itself. Younger users don't search. They scroll. They don't read articles. They consume fragments. The encyclopedia form factor, our twenty-year bet, may be losing relevance faster than any single technology can explain. AI is an accelerant, not the fire.
User:Schiste/what-now, 10 Jan 2026. (Emphasis mine.)

This may also be a major contributor.

I think you're right in that it's probably time we start having uncomfortable conversations as to how to pivot, as there is a chance that the "status quo" is not sustainable for much longer.

The thing is... I don't know if focusing on curation is a good idea, nor do I know if it's a bad idea. What I do know is that it sounds "immediately reasonable", just like 10 other ideas I can think up on the spot. And I feel like we need to do our "due diligance" here.

  • Are other sites (like Reddit, Wikipedia, Quora, etc.) having the same problem? And if not, what's the key difference? Better information should give us a better idea of what needs changing.

  • If we do nothing, does Stack Exchange still exist in 5 years? (Or 1 year?) Or is it simply time to bow out gracefully and try something else entirely?

  • Should we rethink some of Stack Exchange's core ideas?

    • Should we allow duplicates? (Or should "duplicates" work in the other direction, and the newer question remains open?)
    • Should we seek opinions and personal experience?
    • Should we continue using a question and answer format?
    • Should we pivot to entertainment/fun?
    • Should we pay Stack Exchange's content creators? I.e., actual money e.g. through ad revenue; maybe like Medium. In some places, teaching AI is a paid gig, and we're doing it for free.
    • Should Stack Exchange content be made private? (Want access? Then contribute.)
    • Should we allow different AI answers and let them compete against each other for votes?
    • Should we systematically go through each site, generate a list of everything that could be considered on-topic, and write about each one?
    • Should answers be videos instead of text?
    • Should AI Assist generate interaction summaries, made public to help others (and not lock up conversations in a vault)? (This seems more in line with "making the Internet a better place".)
    • Is it simply too risky for the company to invest in AI?
    • Should we rethink copyright, now that these modern AIs are widespread?
    • Should we have an AI "front end", and human "back end"? Like an AI secretary.
    • Should we be making posts with an assumption that the "audience" includes AI?
    • How can we incentivize human contributions, when their contributions will be sold for profit to AI companies, they get 0% of that profit, the vast bulk of the traffic goes to the AI (not here), and the human is rendered redundant once an AI has read their post?
    • Do we need a kind of "humanhood verification system"?

I honestly don't know the answers to these questions: their feasibility involve multiple parties (random people on the Internet; users who make contributions; Stack Exchange as a company) that sometimes have requirements which cannot be simultaneously satisfied.

TL;DR: We probably should investigate multiple avenues.

Improve this answer
5
  • 8
    The point about fragments versus articles is interesting. There has been a downward trend in adult literacy starting around 2017. I’ve always found that the Stack Exchange community tends to value verbose textual explanations. I definitely think one of the avenues that needs to be explored is how to accommodate people who don’t like to read long-form text (even though I am old and frustrated by the trend of information getting stuffed into difficult-to-search/skim video forms). Commented Jan 11 at 13:55
  • 2
    I don't think AI is as strong a reason for drop in questions as we think it is. If you even just reference the graph alone the decline started from 2016 - well before any AI became prominent. COVID brought a spike in 2019/20 (presumably due to staying at home), but barring this the decline has been steady. Also I don't think young people are adverse to research either as this is a drastic drop - the drop in question activity I would think is simply most questions have been asked. A better graph would be question visits over time - then we can assess relevance properly. Commented Jan 12 at 16:45
  • Rethinking the core ideas of Stack Exchange should be done outside of Meta SE / SO because the current system limitations play against this. Commented Jan 12 at 17:04
  • 1
    I recognise a lot of your hypotheticals are just spitballing, but I thought I'd mention that some of them (answers as videos instead of text, AI answers, etc) would absolutely turn me away from all parts of the network except for chat. Commented Jan 14 at 0:05
  • Yes, the decline started before AI. However, the decline from almost 300k to 150k was survivable. Immediately after ChatGPT released, that went from 150k to 100k in about six months. Since then, it has kept dropping and is now something like 5% of peak. I agree that there are other problems and the other problems mostly showed first, but AI has grown to the largest single problem. I'm not convinced that "all questions have been asked". Reddit and Discord seem to be fine. Switching gears, I would also find many of the proposed changes here offputting. Commented Jan 15 at 16:58
4

SE has already changed its focus, at least the company, now it's focusing on engagement.

This can be seen mostly on Stack Overflow

  1. Comments. Ref. New comment UI experiment graduation
    • The reputation required to comment was lowered to 1.
    • Threaded replies
  2. Open-ended questions Ref. Opinion-based questions alpha experiment on Stack Overflow
  3. It was announced that this month, the reputation required to participate in chat will be lowered to 1. Ref. Stack Overflow chat opening up to all users in January; Stack Exchange chat later
  4. Free Upvotes. Ref. Stack Overflow Experiment: Safely expanding voting access
  5. Discussions were repurposed as Challenges. Ref. New and improved coding challenges
  6. AI Assist. Ref. AI Assist is now available on Stack Overflow

Also, they have announced a curation initiative with two major components: "The Workshop" and "The Archive". Ref. Modernizing curation: A proposal for The Workshop and The Archive

Expect that some of the above initiatives, such as opening the chat to all users, will eventually be rolled out to the other SE sites.

Improve this answer
1
  • 1
    I agree that this is what they are doing. I disagree that it is the best use of the site. The people who would be attracted by this have already left. The people still here are exactly those put off by changes like this. I would make a new site that works like this and see if Stack Overflow can drive traffic to it. However, good job of summarizing what they are actually doing. Commented Jan 15 at 17:06
-2

There are only so many questions one can ask, even with new technologies/versions coming out.

I disagree, with the explosion of technologies/versions, there should be way more questions every year. The issue for Stack Exchange is that AI can satisfyingly answer most of these questions nowadays.

I agree that improving curation wouldn't hurt though, e.g. allowing minor edits that don't necessarily bump questions into main page or better versioning.

Improve this answer
29
  • 1
    Of course there will be new questions, but the rate of new questions once the "basic" questions have been answered will be dramatically lower. E.g. You don't need to ask why sorted arrays are faster for every new JS version. As for AI answering then sure it can so long as someone else already did and the AI was trained on it., ideally the AI would be trained form SE content. Commented Jan 7 at 19:18
  • 2
    @B-Tech Many questions on SO are just RTFM, which is easy for AI since it did RTFM. Agreed for basic language questions but there are tons of new libs/techs. You're maybe correct for some SE sites whose themes have little knowledge expansion. Commented Jan 7 at 19:22
  • 18
    Rather, the problem is that people think that AI can answer them. The average GenAI out there doesn't have a clue about new technologies released in the past year because it's training data will very likely be too old. At best it will try to look it up on Reddit where people don't have a clue either. Commented Jan 8 at 8:42
  • 2
    Except that's what for example SO tried and the result turned out even worse than the original AI... Commented Jan 8 at 11:59
  • 1
    @Lundin I only tried the SO's AI assist on day 1 and the retriever had clearly some issue. I already reported 10 others issues/feedback on the AI assist and didn't feel like spending more time to report examples of retriever issues. But you can try Gemini, Claude or GPT with web search on and it's pretty good. There is really no point in using SO's AI assist. Commented Jan 8 at 12:49
  • 1
    @FranckDernoncourt Yeah but like I said, those various GenAI go to poke Reddit and various other crap sites so it's likely to give results of dubious quality. Particularly when it comes to recent events and brand new technology. Commented Jan 8 at 12:55
  • 1
    @Lundin Often go to the official documentation pages. Commented Jan 8 at 12:57
  • 2
    It only works if there happens to be a draft standard that's mostly similar to the actual standard. The deal with GenAI in general is: what you pay is what you get. Commented Jan 8 at 13:01
  • 1
    @Lundin Most of the information needed to answer these questions is public. Also you've shifted from the knowledge cutoff date issue to the paywall issue. I don't really want to play OpenAI customer service here by addressing all the typical customer complains... the reality now is that tons of people have left SE, people use AI, AI firms make $$$, the number of questions on SO has decreased 25-fold (2k new questions per week on SO vs. billions of messages to ChatGPT daily), many SE sites are now deserted and benchmarks show that AI are strong at Q&A. Commented Jan 8 at 13:48
  • 8
    The very real issue I faced with ChatGPT is that it cannot answer questions about C programming correctly, since the latest standard is very different from the draft which ChatGPT has soaked up indirectly from Reddit or somewhere. Instead it lies. With enough people quoting the actual standard, the AI might turn correct eventually, but that process will likely take years. Similarly, when the standard was new it was also lying because it picked up incorrect info from older draft versions. So it can neither handle new info nor published ("paywall") info. Commented Jan 8 at 13:55
  • 1
    @FranckDernoncourt most people looking for help don't have access to the pay walled docs either. And even if they did, trying to dump a 500 page pdf as context isn't nessicarily going to work. Commented Jan 11 at 13:52
  • 1
    @Lundin oh, you think that human dev is correct 100% of time? XD Vast majority of software is dripping with bugs. Most of them are completely irrelevant (if your app crashes from time to time, then just restart it). And those few critical won't be solved by SO anyway. Yeah, I'd rather have AI correct 80% of time, which saves 3/4 of my time. Commented yesterday
  • 1
    And there's another thing. Perhaps even more important. Do you know how unfriendly SO "community" is? Especially moderators. I despise those people. Whenever I ask a question someone always comes and tries to "teach me" how to behave, how to ask questions, or whatever. Instead of answering the question and actually helping someone. A bunch of self-righteous megalomaniacs. Fkin joke. AI is always nice, polite and friendly. Commented yesterday
  • 1
    @freakish Many people share your view: meta.stackexchange.com/a/408833/178179 Commented yesterday
  • 1
    @FranckDernoncourt yeah, I'm sorry that your answer is so downvoted. Its an irony to be honest. All I can do is give it +1. Thanks for the support. Commented yesterday

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.