Feedback Guidance

Some aspects of feedback to SmokeDetector have historically caused confusion. This page is intended to address some of that confusion.

For guidance on when or why to watch or blacklist something, see Guidance for Blacklisting and Watching.

Generic Guidance

In very general terms, the litmus test for whether you should use k or f is this:

If SmokeDetector was implemented as a system-level block, would we want to catch this type of activity?

If the answer to that question is yes, you should mark the post k. Otherwise, a f or n response may be more appropriate.

If it’s not obvious which type of feedback was appropriate for a post, it may be easier to refrain from giving feedback until you can discuss it with others in Charcoal HQ and work out what the correct response is. You can also leave a comment on the metasmoke record to help others looking at the record work out what’s already been thought of. Please don’t comment on the post itself on Stack Exchange unless you’re a regular of the site.

There are a few types of activity we have specific feedback guidelines for, as outlined below.

Disclosed affiliation

It’s fine to promote your own product or service on Stack Exchange, as long as:

If all of those conditions are true, then self-promotion is not spam and therefore f is the appropriate feedback, not k (unless it doesn’t answer the question, in which case it’s n). If any of these conditions are false, self-promotion is spam and, therefore, k is the appropriate feedback.


Self-vandalism is where a user vandalises their own post by replacing all its useful content with something like “xxxxxxxxxxxxxxxxxxxxxxxx”, or “deleted deleted deleted”. At a system level, we want to catch and correct this type of activity. Therefore, for self-vandalism, use tp- (or one of its aliases, such as v, vand or vandalism).

If the user has vandalized the post more than once and/or more than one post, then you should raise an “in need of moderator intervention” flag explaining that the user is vandalizing their posts and/or persistently vandalizing a single post.

If you have full editing privileges on the Stack Exchange site where the vandalism occurred, then when handling reports of self-vandalism, in addition to giving tp feedback to SD, you should also roll back the vandalizing edit and leave a comment warning the user not to do so again. For example:

Do not vandalize your posts. By posting on this site, you've irrevocably granted the Stack Exchange network the right to distribute that content under the [CC BY-SA 4.0 license](// for as long as it sees fit to do so. For alternatives to deletion, see: [I've thought better of my question; can I delete it?](/help/what-to-do-instead-of-deleting-question)

We hope most users, when warned, won’t try to vandalize their posts again, making it a one-time incident.

If you do not have familiarity with the Stack Exchange site on which the vandalism occurred from activity other than responding to SmokeDetector reports, then you should not leave the above comment, as Charcoal maintains a report-handling policy that Charcoal members should not comment on posts reported by SmokeDetector unless they are familiar with that Stack Exchange site from activities other than those related to Charcoal.

Other than rolling back the vandalism and leaving a comment like the one shown above, there’s really nothing that non-moderators can do to stop the user from vandalizing their posts. If the user continues, then in order to resolve the situation, a moderator needs to step in and either suspend the user and/or lock the post(s). You should not get into a rollback war with the user by repeatedly reverting vandalism on the same post. If the vandalism is rolled back once or twice and the vandal continues to vandalize a post, the situation should be left to moderators as normal users do not have the abilities necessary to resolve it.

The feedback aliases v, vand, and vandalism do not automatically add the user to the blacklist. In general, you should not add the user to the user blacklist. There isn’t much benefit to having every post by the user be reported, and there are drawbacks. If the user is added to the user blacklist, then their posts will be reported for both the vandalism edits and the rollbacks to restore the original version. That tends to be disruptive and somewhat hides the actual vandalism reports. So, in general, don’t put vandals on the user blacklist. There are times when adding the user to the user blacklist is reasonable. Usually, that’s when the content they are replacing their posts with is rude/abusive, which should be reverted more rapidly if possible.

Posts containing only garbage

For posts containing no understandable content (e.g., “dfajiojaifojadiofjadhiga”), these should be given tp or tpu feedback, because we would want a system-level block of this content.

However, if you choose to raise a flag, you want to be a bit more nuanced about which flag you use. Shog9’s answer on MSE indicates that, in the vast majority of cases, almost any flag works, but that he’s partial to “rude or abusive” (R/A), as those flags get the post deleted faster. For users with more than a very small amount of reputation, you should stay away from red-flags (spam or R/A) (MSE). Instead, give the user the benefit of the doubt by raising a “not an answer” (NAA) flag. In addition, be aware that which flags are acceptable for this type of content are different on some individual sites. On some sites, flags other than “rude or abusive” and NAA may be declined (e.g., some sites will decline spam flags on this type of content: physics).

Foreign-language posts on English sites (or vice versa)

Most sites on Stack Exchange are English-only, but some permit (or require) other languages. Posts which are in a language other than those used on the site where they were posted are handled largely based on their translation into the appropriate language, with the exception that answers are either NAA or TP, depending on content. Questions are FP and answers are NAA, unless a translation indicates the post contains text which would be considered TP (i.e., if it’s spam or R/A, the correct feedback is k). Being in the wrong language for the site alone doesn’t make a post a true positive. Typically, most of these can be determined by using automated translation (e.g., Google Translate). However, it’s sometimes necessary to get the opinion of a native speaker. We have a list of people who have indicated they are willing and able to review posts in some languages other than English.


Plagiarism is copying without attribution from another source such as someone else’s answer, another website, or a book. If you get to a post before it’s deleted, an easy way to spot some plagiarism is to check the other posts on the same page for similar content. But that won’t catch everything. In many cases, plagiarism is difficult to spot.

If you notice that a post is plagiarized, then mark it as tp. You may wish to add a comment along with your feedback indicating that the post is plagiarism (and giving the original source) in order to make it clear why you’ve given tp feedback and avoid arguments about what the feedback should be.

Note that we’re not aiming to catch plagiarism. Noticing that a post that is neither spam nor rude/abusive (R/A) happens to have been plagiarized is a nice bonus, but plagiarism is not something that SmokeDetector is intended to catch. Therefore, while plagiarism by itself should be marked as tp, the !!/report command shouldn’t be used to report plagiarism and the user should normally not be placed on the user blacklist. (It’s possible for there to be additional circumstances which make putting the user on the user blacklist [i.e., tpu] and red-flagging appropriate [e.g., a persistent troll who repeatedly posts inappropriate plagiarized content; there are, unfortunately, some trolls like this], but that’s not the normal response to just a single plagiarized post.)

As to flagging, plagiarism is usually handled by raising an “in need of moderator intervention” flag in which you explain that it’s plagiarism and provide a link to the original source (or at least explain why you think it’s plagiarism). Plagiarism on its own is not normally handled with either spam or R/A flags.

AI-generated content (e.g., ChatGPT)

Content that is AI-generated is considered “the work of others”. If AI generated content is not properly indicated (i.e., in block quotes) and attributed in the post according to the referencing requirements, then the post should be handled just like plagiarism.

Note that some spammers are now using ChatGPT and similar tools to hide their spam in posts that look superficially plausible. Such posts are spam, independent of whether they are AI-generated, and should be marked as tp. If you encounter such a post which is not reported, it should be reported (using the !!/report command). In every respect, spam should be treated and handled as spam, regardless of its origins, form, or the content it’s in.

However, if the post is not spam or rude/abusive, but just AI-generated content, then look to see whether the AI-generated content is properly referenced as per the referencing requirements and the official policy from staff in response to the question “Is attribution required for machine-generated text when posting on Stack Exchange?”. If the AI-generated content is not in a block quote and properly attributed, then it’s effectively plagiarism, in violation of network-wide policy, and, thus, should be marked as tp, but not manually reported when not automatically detected. Otherwise, properly-referenced AI generated content is either NAA (n) or false-positive (fp).

Independent of the feedback that you give to SmokeDetector, you may choose to raise an “in need of moderator intervention” flag on posts you believe were generated by an AI, explaining your suspicions (and presenting any evidence you can offer to justify them) to a site moderator. However, the policies for handling AI generated content, other than the need to follow the referencing requirements, are made on a site-by-site basis. If you want to see what a particular site’s policy is with respect to AI-generated content, please see this community-maintained list of ChatGPT discussions and policies.

Repairable offensive posts

Posts that contain offensive language (e.g., expletives) are true positives if they cannot be salvaged by reasonable editing within Stack Exchange’s normal editing guidelines, or if the user is using language which appears to be intended to be offensive. If it’s a Code of Conduct violation, then tp feedback is correct, even if the post is repairable.

An example of a repairable post is that some people use f*cking as an emphatic synonym for very. Those posts usually can be salvaged by editing out the inappropriate language. In such cases, the appropriate feedback is f, as we don’t want a system-level block preventing these kind of posts. The word “salvaged” here means there would be something of value to the site after the offensive portion(s) were removed/rephrased. Yes, this guideline means posts that contain some minor offensive language are fp.

If there’s no content to salvage, then tp- is appropriate—or tpu-, if you feel the post indicates the user may repeat similar content, such that we would want to block them at the system level.

Note that the rules are different when a post is edited by its author in a way that is offensive; such cases should be handled as self-vandalism. Basically, vandalism of a post that adds offensive language is tp-, as we would’ve wanted that edit to be blocked by the system.

NAA feedback

NAA feedback has sometimes been a little confusing. Using n as your feedback should, in general, be done when:

Bad questions

NAA feedback is disabled on questions, as questions are, in fact, not answers. But, what if the question is VLQ, off-topic, too broad, or otherwise considered ‘bad’ by the site’s quality standards? Unless the question is actually spam or rude/abusive, then the appropriate feedback is fp-. Charcoal is not intended to be the arbiter of what’s on-topic for any particular site. It is possible for such posts to go over the line to the point that they can/should be marked as TP, but that should be based on a spam/rude/abusive spectrum, rather than topicality.

If you’re still unsure, err on the side of caution by choosing f. Then, leave a comment on the MS report or in chat indicating what is wrong with it, and why you chose that feedback.

User expressing suicidal intent

Please avoid raising red flags on such posts, unless the post is far enough into the spam/rude/abusive spectrum such that it’s significantly inappropriate for others to be reading/for it to be displayed. Such posts should be handled by raising an “in need of moderator intervention” flag on the post to bring it to a moderator’s attention. A moderator on the site should raise a Community Manager (CM) escalation, for which there is a specific category. Such escalations receive priority processing by the CMs. The desire for red flags not to be raised is because the effects of the red flags may be yet one more negative thing for the person, who is already in a vulnerable place. However, that shouldn’t mean that all someone needs to do to not get red flags for a post that’s truly offensive is to say they are considering suicide. So, use your judgment, but please lean fairly strongly towards not raising red flags and allowing moderators/CMs to handle the situation.

Feedback, on the other hand, should be based on a normal evaluation of the post (i.e., if it’s spam or R/A, then it’s TP).

Posts which you manually report

It is preferred that you not provide TP feedback on a post that you manually reported. We want another pair of eyes to review your reports. While this isn’t as critical as it once was, due to the required number of feedbacks having been changed from 1 to 2, it’s still expected that you not give TP feedback on your own reports. This is particularly true where the only reason for detection (i.e., the why data) is that the post was manually reported.

Note that, if you report something in error and/or later change your mind, it is reasonable for you to provide FP or NAA feedback on that report, even though you manually reported it.