57

Recently, I observed there is a steady stream of code-only answers, which seems completely irrelevant to the question.

A few examples:

  1. https://stackoverflow.com/a/79557842
  2. https://stackoverflow.com/a/79557285
  3. https://stackoverflow.com/a/79557104
  4. https://stackoverflow.com/a/79547672
  5. https://stackoverflow.com/a/79548401
  6. https://stackoverflow.com/a/79547833
  7. https://stackoverflow.com/a/79559335
  8. https://stackoverflow.com/a/79559340

My observations:

  1. Most of the content is just some code dump that is completely irrelevant to the question
  2. Most of them are created by 1-rep accounts
  3. Some of them involve non-English content

Currently, what I am doing for most of them is to raise an NAA flag and leave a comment to state my observation that it is not relevant to the question. I have also tried to raise R/A flags on some of them and all of the R/A flags are marked as helpful.

However, I am not sure where the boundary is and I hope I can have the community's suggestion(s)/clarification(s) on what is the best action(s) I should take on this kind of post, in order to stop this non-sense wave?

Edit: I humbly disagree with the dupe vote to Reviewing Low Quality Posts - Answers without explanation. From my perspective, my question focus on the blatantly nonsense nature of the answers; while the dupe source focus on the case that "may or may not answer" the question. As pointed out in comments, this could be speculation, but these posts might be AI-generated spam seed/special kind of DDOS. I hope this post can raise general awareness on the issue.

Credit to gnat's comment for formulating this difference!

53
  • 4
    You have enough rep to cast a delete vote on answers with negative scores; that's a good option for many cases like this. NAA may be appropriate sometimes, but reviewers may not always be able to discern whether a code-only answer is off-topic garbage or a genuine attempt to answer. Commented Apr 7, 2025 at 2:53
  • 14
    @adabsurdum Thanks for the suggestion first. I think casting a delete vote on answers is a privilege that unlocks at 20k rep. Guess I need to work more on that :) Commented Apr 7, 2025 at 3:11
  • 6
    I'd say no. My R/A flags on Q&As that were VLQ and included some random characters to bypass the filters are getting rejected recently. I remember them marked as helpful a couple of years ago (memory can trick us). So, either the consensus in general shifted regarding handling R/A flags or some moderators feel those flags and their penalty should not be imposed on such posts. Examples that you've included are far less abusive, imo. I'd personally dv and vtd, so, as I said, I don't think you should flag these as R/A unless that's a recurring issue for a user which then should be mod-flagged. Commented Apr 7, 2025 at 7:02
  • 4
    "blatantly non-sense" doesn't really apply. I'm not a SME, and the answers look relevant enough to the question. Note that mods are generally not SMEs in any particular language/tech. Also, "and all of the R/A flags are marked as helpful" is probably because the answer was deleted with a pending R/A flag on them. Commented Apr 7, 2025 at 7:59
  • 6
    Anyone will say something about the fact that most of these answers are on necromantically raised questions, from 2010-2012 years? There is a flood of these as well. Even while I have a Necromancer badge, I didn't do that on purpose, and my answer was accepted. Commented Apr 7, 2025 at 8:50
  • 12
    Honestly this looks like an attempt at poor man's DOS attack, flooding the site with AI-generated non-English spam answers. I don't have the tools a moderator would to check if this is a new thing or if it has always happened this frequently. Something like this has not been reported in the past. Commented Apr 7, 2025 at 14:12
  • 4
    @ThomA formally, one can agrue that code with comments does not qualify as code-only anymore. And most importantly, if we let them slip through, how long do you think it will take for abusers to learn that code comments make a reliable disguise for their trash\ Commented Apr 7, 2025 at 14:22
  • 6
    I've just been flagging them NAA but I am concerned they're up to something. Creating legitimate-looking accounts for future use somehow? It doesn't make sense for a regular human to do this. Commented Apr 7, 2025 at 16:26
  • 3
    This question is similar to: Reviewing Low Quality Posts - Answers without explanation. If you believe it’s different, please edit the question, make it clear how it’s different and/or how the answers on that question are not helpful for your problem. Commented Apr 7, 2025 at 17:32
  • 4
    Related: In the past month, Translate Java Code into PseudoCode got three Java-code-only answers that were totally unrelated to m68k assembly or pseudocode for if/else, and just seemed to be implementing some random small example program; it's not plausible these were good-faith efforts by a human to answer the question, not even if they'd used ChatGPT. They did use code blocks correctly, unlike some you linked, but smells similar. The question already had an accepted answer from 2012. (Which yesterday got its first upvote...) Commented Apr 7, 2025 at 18:12
  • 7
    I've had a declined flag ~half a year ago on exactly this kind of post - badly formatted code dump from a 1-rep user that has nothing to do with the question (the code in answer seems to send emails from python code, the 10 y.o. question is about editor autocomplete behavior), upvoted very quickly, somehow passing both Late Answers and First Answers queues. Likely just some misunderstanding, but the answer still exists... stackoverflow.com/a/79229072/14401160 Commented Apr 7, 2025 at 19:43
  • 5
    this is different from Reviewing Low Quality Posts - Answers without explanation because answers discussed here don't qualify as ones "that may or may not answer the question". Quite the oppposite, these clearly and blatantly don't answer the questions and don't even indicate an attempt at that (consider editing this into your question to make it easier for readers to see the difference) Commented Apr 8, 2025 at 7:22
  • 7
    Wow, the pdf tag seems to be rank with these sorts of nonsense answers. Post after post that create some hardcoded PDF in python that has nothing to do with the question at hand. The earliest I've found so far is this one from Feb 28: Reportlab alignment & width issues. Commented Apr 17, 2025 at 19:35
  • 6
    I recently saw an enlightening comment under such "answer" (10K link): "could you keep it up for another 3 hours at least please". Looks like poster (and maybe others) use answers only to share random code snippets - of course totally ignoring the question. I think this is a delete-worthy abuse... Commented May 29, 2025 at 16:24
  • 7
    @gnat another evidence to support your theory of personal pastebin: stackoverflow.com/a/79646806 Commented May 31, 2025 at 18:02

3 Answers 3

Reset to default
30

tl;dr

Flag the post as "Looked at by a moderator" and include a link to this Meta post in the flag text.

The usual advice about how to deal with technically irrelevant posts provided in this answer is spot on, and normally I would recommend to follow it without raising "Looked at by a moderator" flags. Community moderation would suffice.

Yet in this particular case:

  1. The answers are unsuitable for review queues because code-only answers tend to sneak through and we can't blame the reviewers for not knowing about this Meta post.

  2. It's useful to build up a paper trail across all these accounts in the form of custom flags.

To get the post removed on the grounds that it is indeed not an attempt to answer the question, please follow Dharman's advice (quote amended by me):

I'd just like to point out that if you use NAA/VLQ flag it will most likely be declined. Mods will not know that you used this flag because the code doesn't answer the question. Better explain the issue in your own words using "[I]n need[] of moderator [intervention]"

June 11th, 2025

Currently, the most reasonable theory, also mentioned in the comments to this post, is that some folks are trying to use StackOverflow as a notepad, mainly for cheating on course exams or whatnot. It's easy to imagine how a CS course may allow using Stack Overflow or similar websites during an exam or to complete an assignment. Answer boxes are free text fields, plus we permit unregistered users to post answers. There you go.

I think this calls for outright destroying the accounts as the most expedient moderation measure. This behavior amounts to an abuse of the system.

8
  • Would it be of help to flag older deleted code-only answers that follow the pattern described here? I specifically mean ones deleted via regular delvotes. Commented May 15, 2025 at 8:45
  • 2
    @VLAZ no, mainly because regular users don't know if a post was already flagged and how before it got deleted via community vote. If you mean older than this MSO question, still no because there's no guarantee that the issues are even connected. The risk is to produce more noise than signal. Commented May 15, 2025 at 8:48
  • 8
    by the way, one of these posters recently leaked a comment suggesting that they use SO simply as a pastebin service, "could you keep it up for another 3 hours at least please". I hesitate to recommend regular users flag these cases as R/A but for moderators it probably makes sense to delete such posts as abusive in order to throttle vandalism Commented May 30, 2025 at 6:17
  • 3
    "for another three hours", but for whom? maybe it's once more a "class assignment" situation (someone has to witness the post or post count). doesn't change the response it should receive. immediate deletion and perhaps temporary suspension. Commented May 30, 2025 at 10:39
  • 3
    @ChristophRackwitz Doesn't seem like suspension would have any effect in these cases, since a new user is created for each answer, and in every case I've seen, never returns after posting. Commented Jun 10, 2025 at 19:58
  • 3
    And here's a question that has attracted six nonsense answers (seven if you include a deleted answer, and eight if you include a borderline answer): How to create a PowerPoint with Excel VBA without seeing the application. Honestly, WTF? Commented Jun 15, 2025 at 21:25
  • 2
    Don't the user have access to their deleted answers? In which case, deleting the answers is not really defeating the "pastebin" thing, is it? Commented Jul 25, 2025 at 16:18
  • 5
    @chrslg yes they do (if logged in), that's why destroying abusive accounts seems to be more reliable way to handle this. I think it is only when account is destroyed they really lose any way to see deleted posts Commented Jul 25, 2025 at 16:21
23

Code-only answers are discouraged and low quality. To determine if you should do anything further except down-voting, you have to study the answer to see if it tries to answer the question. Answers which doesn't answer the question are not an answer/very low quality and can be flagged as either of those reasons.

For example if the answer is in a different programming language than any ever stated by the question, it is off-topic. Check question edit history to ensure that it didn't ask about said language at some point in time, in which case the answer might be OK.

The answers you have identified all seem quite fishy. Not only do they not answer the question, they could be "spam seeds" (posted as "sleeper agents" in order to later be edited & replaced with spam), maybe also GenAI generated. So you did the right thing posting this on Meta, thereby giving the moderators a heads up.

I'd suggest you deal with all such answers in the following way:

  • Check so that it isn't a legit answer by a confused user, trying to answer (any historical version of) the question.
  • Check so that there are no embedded URLs to fishy sites - then it is probably outright spam and should be flagged as spam.
  • Down-vote and flag as not an answer/very low quality. Delete vote if you can.
  • Go to the SO Close Vote Reviewers chat and post a line like this:

[tag:del-pls] [tag:20k+] Does not answer the question, possible spam seed. [link to the answer]

High rep users with delete privileges and moderators may then actively deal with the post in a timely fashion, in case there is something fishy going on. You can also report outright spam posts in the chat by using the [tag:flag-pls] [tag:spam] tags.

High rep users (15k+) can also protect posts from getting answers posted by 1 rep users, to guard against things like this. If you see a specific post getting repeatedly targeted, it should get protected. Those who don't yet have the rep to do that can also ask for help with it in the mentioned chat.

11
  • 10
    If you flag such posts as VLQ, please leave a comment explaining why you think it has nothing to do with the question. Don't expect reviewers to find such things as we are not necessarily subject matter experts. It might be easy to see if it is in a different programming language, but code only posts in the correct language are pretty hard to identify when you don't know what to look for. Commented Apr 8, 2025 at 7:34
  • 1
    VLQ should not be used if it's required to explain why is VLQ. LQ posts should be downvoted. Commented Apr 8, 2025 at 13:54
  • 4
    @Wicket answers discussed here are strange beast. I got used to avoid flagging VLQ and recommend deletion in LQ queue at code dumps and I would skip this stuff if there were no comments explaining the issue. However when there are comments suggesting me to just look at the question and compare it to such "answer" things turn into unexpected way. It gets easy to see that answer is just a random text blatantly ignoring what was asked and, which feels really unusual - I can see that even in questions outside of my expertise. In that sense these are ideed close to cases when I'd use canned flag Commented Apr 8, 2025 at 20:14
  • @gnat, AFAIK, comments can't be addressed to reviewers, and it's not expected that reviewers review comments. It's a lot better to flag these strange beasts as R/A. Commented Apr 8, 2025 at 22:00
  • @Wicket my point is not about comments, their role here is merely in making me look at how answer relates to question, nothing more. Weird things start after that - answer that I thought I would simply ignore or maybe vote down if it's in my area of expertise suddenly turns out to be like stuff I tend to be flagging - random text dump having nothing to do with the question Commented Apr 8, 2025 at 22:06
  • ...I maybe could do this without comments if I had a habit to look at questions - which I don't do in cases of flaggable answers. Except for link-only ones which as you probably know require one to check the question to decide whether it's NAA or not Commented Apr 8, 2025 at 22:15
  • 1
    @BDL For example if the question is about Python (and was always about Python), then a code-only answer in C++ is VLQ. There's no real need to explain that in comments, though comments are helpful. And yeah that VLQ review might have to be carried out by users who know either Python and/or C++. Others who are unsure should skip the review. Skipping review is fine and encouraged whenever one is unsure. Commented Apr 9, 2025 at 14:52
  • Rather than leaving a comment below the VLQ answer it might be wiser to go to the mentioned SOCVR chat and leave a "del-pls" request tagged with Python. Then users there who feel confident enough about Python can go ahead and vote. Commented Apr 9, 2025 at 14:55
  • 6
    I'd just like to point out that if you use NAA/VLQ flag it will most likely be declined. Mods will not know that you used this flag because the code doesn't answer the question. Better explain the issue in your own words using "in needs of moderator attention" Commented Apr 17, 2025 at 14:22
  • 1
    @Dharman - Posting del-pls requests in SOCVR as indicated in this answer has been working well for me so far. But I've VLQ-flagged a couple of code-only answers that were full of non-English code comments, do you think that will be a problem? Commented Apr 17, 2025 at 19:40
  • 4
    @dbc There's just a high chance that such VLQ flag will be declined. Commented Apr 17, 2025 at 20:27
10

This has been puzzling me for quite a while. It was particularly concerning that I could not figure a rational explanation for this behavior.

I considered a bunch of old known reasons seen for similar posts in the past. Random mistakes, spam seeds, trolling, DDOSing, ChatGPT, faking proficiency at job interviews. None of these felt good enough. Such answers appear too frequently for random mistakes but too infrequently for DDOS. I have never seen them end with spam edited in. They look too bland for trolling. They look too irrelevant for ChatGPT word salad. They are too unreliable for job interviews - because interested interviewer will likely have a closer look into the answer and discover cheating - especially if gets some -1 while hanging in a resume. Etc

Only theory that came to mind to tick all boxes seems to be this. Some teachers probably incentivise their students for "successful" posting at SO - assuming that this indicates some verifiable proficiency level. Students probably show them their logged in profile with non-negative score answers and get additional grade points, something like that. Some particularly unscrupulous teachers may even accept minor negative score - "oh -1, that's probably some minor mistake, acceptable for a newbie SO user".

If you think of it, dumping random code snippets at random irrelevant questions makes a perfect trick for lazy students of such (equally lazy) teachers. Whole SO system is designed to let it leak through. Code dumps have solid chances to sneak through LA, FA, LQA reviews no matter how irrelevant - because none of usual deletion criteria matches. Voting down is also hardly an issue because given total amount of poor quality answers many (most?) would hesitate to inwaste their hard earned rep into downvoting yet another one.

Students willing to stay on the safe side may even create two accounts and dump two "answers" and show their teacher one that survived long enough at reasonably acceptable score like 0 or maybe -1.

This seems to be pretty much the same thing as we already see in multiple homework dump questions. Lazy students don't know what is a good quality post and neither they care. They just throw their stuff against the wall and see if it sticks.

Next question I asked myself, okay if it's much like homework questions - why does it feel more painful? After all, questions are more visible and there are certainly more homework dumps than such answers (and probably will always be), so what's up?

My guess is, here it is important that question downvotes are free unlike answers.

When I see a homework dump question, I just vote it down (maybe optionally voting close) and go on with my life knowing that system will take care of the rest - lower visibility, asking rate limits, roomba - and it's gone. It costs me just few seconds - somewhat annoying but no big deal really.

With such answers, there is nothing like that. I could spend all my 30 downvotes a day on these and all I get will be just lost 30 rep points. Soon I will lose enough points to get out of voting down privilege (even sooner I will get out of VtC privilege) and will be unable to do even that. Helplessness, I really don't like it.

Granted, approach with pinging 20K users explained in another answer would work, but it takes so much more effort compared to how I handle homework dump questions. And I doubt that it can scale well if (or rather, when) more students will learn to use such tricks.

6
  • 4
    I'm not convinced that they're homework dumps - bad homework-related contributions have tended to come in big bursts (up to the homework deadline), while these answers are spaced out over multiple months. They're also in a whole bunch of different languages (text-languages, not programming languages), which doesn't suggest a single university. With that said, I don't have a better theory. Commented May 14, 2025 at 20:43
  • @DavidW difference in distribution over time looks natural to me. Homeworks tend to have strict deadlines and dumps probably follow these. As for additional activities like described, these hardly have anything like that: I can't imagine a teacher stupid enough to tell their students make a successful SO post in something like a week. Regarding the theory itself, I think you have a point - it is in my opinion merely more plausible than other considered here. I wouldn't be surprised if we eventually discover some better explanation Commented May 14, 2025 at 21:04
  • 3
    Agree with David - AFAICT, these have been going at the rate of around 3-4 per day for over a year now. That doesn't quite fit the homework model. Commented May 14, 2025 at 21:06
  • 3
    FWIW, I think it is someone (somewhat) subtly testing a bot/AI tool, for eventually nefarious purposes such as fraud, spam, or DDoS. That's just my gut, though. It feels like a more sophisticated effort than what we normally see in that it clearly involves patience and persistence. However, I would agree that Occam's razor would indicate it's likely just some academically-motivated cause. Commented May 14, 2025 at 21:09
  • @TylerH now that sounds like a sensible theory, too. I had to discard AI in my initial evaluation because I thought it would consider the question and make more relevant code dumps but now that you pointed, it looks unnecessary. One can use it to generate totally random code ignoring the question and reasonably expect that it will have acceptable "rate of survival" no matter how irrelevant Commented May 14, 2025 at 21:20
  • idky i have similar feeling, unsure to call it DDOsing however these could be generated and posted by some Gen-AI- bots potentially! Commented Jul 27, 2025 at 0:28

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.