X's own automation rules contain one sentence that describes most AI reply tools on the market: "sending automated replies to posts based on keyword searches alone is not permitted." That is the business model of a tool that watches for keywords and posts a reply for you. And X is the lenient one. Hacker News bans AI-edited comments, Stack Overflow bans AI drafts entirely, and Product Hunt's commenting guidelines say "No LLMs or Chrome extensions please!"
I read the current rules on eight platforms plus the FTC's rules on undisclosed endorsements on November 21, 2026, because "will this get my account banned?" is the question to answer before you buy any auto-reply tool. Every rule below links to the page I read. I quote short phrases so you can check the wording yourself.
Disclosure: I build Needle, which finds people asking for a tool like yours and drafts a reply that a person reviews and posts. So I have a view on auto-posting. I've tried to keep this to what the rules say.
In brief
- X allows AI reply bots only with written approval. Its automation rules require "prior written and explicit approval" for any AI reply bot, and ban automated replies triggered by keyword searches.
- Hacker News, Stack Overflow, DEV comments and Product Hunt ban AI-written text. On these four, even a human-edited AI draft breaks the rules. Write it yourself.
- Reddit doesn't ban AI text sitewide, but subreddits do. Reddit's sitewide rules ban spam and "content manipulation", and it now labels accounts that use automation.
- LinkedIn bans automated commenting outright. Its User Agreement prohibits bots that "create, comment on, like, share, or re-share posts."
- Disclosure isn't optional. The FTC says employees promoting their company's products "should disclose your relationship to the company", and a profile mention isn't enough.
The per-platform table (checked Nov 21, 2026)
"Automated posting" means software posting replies without a person pressing the button each time. "AI-written content" means text drafted by a language model, even if a person posts it.
| Platform | Automated replies allowed? | AI-written content allowed? | Disclosure rules | Typical consequence |
|---|---|---|---|---|
| X | Only to users who opted in; keyword-triggered auto-replies banned; AI reply bots need X's written approval | Yes, within the rules (bots need approval) | Optional "automated" account label | Posts filtered from search, reach limits, suspension |
| Automated accounts get an [App] label; spam and content manipulation banned | Not banned sitewide; subreddits set their own rules | Platform: none for AI. Subreddits vary | Content removal, account restrictions, suspension | |
| Hacker News | No: "please don't automate posting" | No, including AI-edited comments | Not addressed | Not spelled out in the guidelines |
| No: bots that comment, like or share are prohibited | Not addressed directly | Not addressed (FTC rules still apply) | Reduced visibility, labels, removal, account restriction | |
| Bluesky | Not addressed directly; spam and manipulation banned | No general AI policy | Not addressed | Restrictions or removal |
| Stack Overflow | n/a | No: AI tools "may not be used to generate content" | n/a | Deletion, lost reputation, warnings, suspension |
| Product Hunt | No: "No LLMs or Chrome extensions" | No (comments) | n/a | Hidden or removed comments, loss of commenting access |
| DEV (dev.to) | No AI comments | Articles yes, with disclosure; comments no | Articles must note AI involvement | Lower visibility, suspension or ban |
The details, platform by platform.
X (Twitter): AI replies need written approval
X's automation rules (updated April 2026) are the most specific of any platform here, and they address AI replies by name.
- Keyword-triggered replies are banned. Automating replies "to reach many users on an unsolicited basis is an abuse of the feature," and replies "based on keyword searches alone" are not permitted.
- Automated replies need opt-in. You can only auto-reply to someone who "requested or have clearly indicated an intent" to hear from you, for example by replying to your post or sending you a DM. X adds that someone following you is not enough.
- AI reply bots need approval first. X allows AI-powered reply bots, but "the deployment or operation of any AI reply bot requires prior written and explicit approval from X."
- Scripting the website is a suspension risk. X lists "non-API-based forms of automation, such as scripting the X website" as something that "may result in the permanent suspension of your account." That covers browser extensions and headless browsers posting for you.
- You own what your tools do. The page tells users: "You are ultimately responsible for the actions taken with your account."
X's authenticity policy adds the spam side. It bans "bulk, aggressive, high-volume unsolicited replies," "promoting content by replying with content that is irrelevant to the topic of the original post," and copy-paste replies ("Copypasta"). Enforcement runs from anti-spam challenges and restricted reach (excluding posts from search, downranking in replies) to permanent suspension.
X does offer a label for automated accounts, which "allows automated accounts to identify themselves." A label doesn't make keyword-triggered replies acceptable, though.
Reddit: no sitewide AI ban, but subreddits decide
Reddit's sitewide rules don't mention AI. Rule 2 says to "post authentic content into communities where you have a personal interest," and bans "content manipulation (including spamming, vote manipulation, ban evasion, or subscriber fraud)." Rule 5 bans impersonating "an individual or an entity in a misleading or deceptive manner." The enforcement list runs from "asking you nicely to knock it off" to permanent suspension. (Reddit's own pages didn't load for me, so this wording comes from the copy of the Reddit Rules archived in the US Supreme Court's record in 2024. The live page is redditinc.com/policies/reddit-rules.)
What changed in 2026 is automation. In March, Reddit announced that accounts using automation will carry an "[App] label so users can recognize that they are interacting with software" (Help Net Security), and that accounts suspected of being bots can be asked to prove a human is behind them (TechCrunch). TechCrunch reports that Reddit doesn't prohibit posts or comments written with AI tools, but moderators can set their own rules.
Those community rules matter more than the sitewide ones. A CHI 2025 study of more than 300,000 subreddits found that AI rules were still "relatively uncommon" but that the number of subreddits with them more than doubled in a year, with larger communities more likely to have one. Self-promotion rules are far more common than AI rules. Read the sidebar before you post, every time.
I've written more about Reddit specifically in how to find customers on Reddit without getting banned and the Reddit shadowban guide.
Hacker News: no generated or AI-edited text
The Hacker News guidelines are short and blunt:
- Comments: "Don't post generated text or AI-edited text."
- Submissions: "Please don't put generated text in HN posts," followed by "please don't automate posting."
- Promotion: "Please don't use HN primarily for promotion." Posting your own work "part of the time" is fine.
- No vote-seeking: "Don't solicit upvotes, comments, or submissions."
"AI-edited" is the important word. A draft from any AI tool that you then polish is still AI-edited text. On HN, use a tool to find the thread if you like, then write the comment from scratch.
The Show HN rules add: "Please don't ask friends to upvote or comment." Pitching your product in someone else's Show HN thread isn't banned in so many words, but it is promotion in a thread about their work, and HN readers notice.
LinkedIn: automated comments are prohibited
LinkedIn's User Agreement (effective November 3, 2025) prohibits using "bots or other unauthorized automated methods" to "create, comment on, like, share, or re-share posts, or otherwise drive inauthentic engagement." It separately bans software and "browser plugins and add-ons" used to scrape or copy the service.
The Professional Community Policies add "Do not spam members or the platform" and ask members not to "artificially increase engagement" with their content. They don't address AI-written text directly. Enforcement ranges from limiting visibility or labelling content to account restriction for "repeated or egregious offenses."
So on LinkedIn, the problem isn't AI text. It's software posting comments. A comment you wrote with help and posted yourself is a different thing from a tool commenting on 50 posts an hour.
Stack Overflow: AI drafts are banned outright
Stack Overflow has the strictest rule of all. Its generative AI policy says AI tools "may not be used to generate content for Stack Overflow," and defines that content as anything crafted "in part or in whole" by such a tool. It goes further: "Please do not draft content for Stack Overflow using large language model (LLM) services."
The consequences: content judged AI-written "will likely be deleted, along with any reputation earned from it," with warnings and possible suspension for repeat cases. If you answer questions on Stack Overflow, write the answer yourself.
Bluesky, Product Hunt and DEV
Bluesky's community guidelines (last updated September 19, 2025) don't have a bot or AI section. They ban spam that disrupts "normal conversations or service use" and efforts to "artificially manipulate features or social signals to gain unearned reach." Persistent disruptive behavior "may face restrictions or removal."
Product Hunt's commenting guidelines say the site "is about person-to-person interactions" and ask for "No LLMs or Chrome extensions please!" They also ask you to avoid "overly promotional, unrelated, or generic comments." Flagged comments may be hidden, and repeat violations can mean "permanently removing contribution access."
DEV allows AI-assisted articles if you disclose it, but its AI guidelines ask members to "not use bots or AI to generate comments on posts" (grammar and translation tools are exempt). AI-assisted articles also shouldn't "promote any business, program, or course (including your own)." Violations can bring "suspension or a ban."
The FTC: disclosing that it's your product
If you're in the US or selling to US customers, the platform rules aren't the only rules. When you recommend your own product in a thread, you are endorsing it, and the FTC cares whether readers know you have a stake.
- Employees and founders must disclose. The FTC's Endorsement Guides Q&A says employees who mention their company's products "should disclose your relationship to the company." It adds: "Listing your employer on your profile page isn't enough."
- Put it where people will see it. The disclosure "should catch users' attention." For text posts, the FTC's Consumer Reviews and Testimonials Rule Q&A says a disclosure "could work at the beginning of a text-only post." A disclosure readers have to click or hover to see doesn't count.
- Fake reviews now carry civil penalties. The Consumer Reviews and Testimonials Rule (16 CFR Part 465) "went into effect on October 21, 2024," bans fake or false reviews and testimonials, covers fake social signals "generated by bots," and authorizes "civil penalties for knowing violations."
The practical version: "I work on X" or "I built X" in the same sentence as the recommendation, not buried at the end. I'm not a lawyer, and this isn't legal advice. If you run paid or affiliate campaigns, get real advice.
What gets accounts banned: the patterns, grounded in the rules
Put the rules side by side and the same behaviors come up again and again. These are the patterns that the rules above name, not guesses about algorithms.
- Replying because a keyword matched. X bans it in exactly those words. Elsewhere it shows up as irrelevant replies, which X's authenticity policy and Product Hunt's guidelines both call out.
- The same reply, many times. X bans "duplicative or substantially similar posts" and "Copypasta." Reddit treats it as spamming under Rule 2. Templates are fine as a starting point; posting the same text in ten threads isn't.
- Volume. "Bulk, aggressive, high-volume unsolicited replies" is X's phrase. Reddit removes bots and spam at around 100,000 accounts a day, according to Help Net Security.
- Software doing the clicking. X warns that scripting its website can mean permanent suspension. LinkedIn bans bots that comment and browser add-ons that scrape. Product Hunt asks for no Chrome extensions. If a tool needs your session cookie or a browser extension to post for you, check the platform's rules before connecting it.
- Sock puppets and coordinated praise. X bans multiple accounts that "engage with the same posts" or "misuse the mention/reply feature." Reddit bans vote manipulation. HN and LinkedIn both ban arranging engagement.
- Hiding that it's yours. Reddit Rule 5 bans misleading impersonation. The FTC expects disclosure. Communities treat an undisclosed founder recommendation as the worst kind of spam once someone checks your profile.
- AI text where it's banned. HN, Stack Overflow, Product Hunt and DEV comments all ban it, whatever you do to the draft afterwards.
None of these say "don't mention your product." Most communities accept a relevant, disclosed recommendation in a thread that asks for one. The bans land on automation, volume, repetition and hiding who you are.
A workflow that stays inside the rules
This is the workflow I'd defend to any moderator on any of these platforms.
- Find the conversation. Search for people asking for what you sell: "any recommendations for," "alternative to," "how do you handle." Do it by hand with each site's search, or use a tool that only finds threads. Finding doesn't post anything, so no platform rule above restricts it.
- A person reads the whole thread. Check what was asked, what others already answered, the subreddit or community rules, and whether the thread is someone's launch. If your product doesn't fit, move on.
- Draft. Answer the actual question first. Mention your product only if it fits, once, with what it does and where it falls short. On HN, Stack Overflow, DEV and Product Hunt, skip AI drafting and write it yourself.
- A person edits and posts, from their own account, with disclosure. Rewrite the draft in your own words, cut anything you can't stand behind, add "I work on X" next to the mention, and post it by hand. No scheduled posting, no extensions, no second accounts.
- Stay for the follow-up. Answer replies. A bot can't do this part, and it's where most of the goodwill comes from.
If you want starting points for step 3, the community reply templates guide has examples by platform.
Where Needle fits: it handles steps 1 and 3 across 14 networks and leaves step 4 to you. It never posts, and its drafts include an "I work on X" line when they name your product. On the four sites that ban AI text, treat any draft, ours included, as notes rather than a reply. If you're comparing specific tools, I've written up ReplyAgent alternatives, a ReplyAgent comparison and a ReplyGuy comparison. Auto-posting tools win on time saved per reply; the question is whether the platforms you post on allow it, and this table is the answer.
Evidence notes
All pages opened on November 21, 2026 unless noted.
- X automation rules (updated April 2026): https://help.x.com/en/rules-and-policies/x-automation
- X authenticity policy (platform manipulation and spam; page dated April 2025): https://help.x.com/en/rules-and-policies/authenticity
- X automated account labels: https://help.x.com/en/using-x/automated-account-labels
- Reddit Rules, quoted from the copy archived in the US Supreme Court record (PDF created 2024; Reddit's own policy and help pages did not load for me on Nov 21, 2026): https://www.supremecourt.gov/opinions/URLs_Cited/OT2023/22-277/22-277-8.pdf. Live page: https://redditinc.com/policies/reddit-rules
- Reddit [App] label and human verification, March 2026: https://www.helpnetsecurity.com/2026/03/26/reddit-human-verification-changes/ and https://techcrunch.com/2026/03/25/reddit-bots-new-human-verification-requirements
- Lloyd et al., "AI Rules? Characterizing Reddit Community Policies Towards AI-Generated Content," CHI 2025: https://arxiv.org/abs/2410.11698
- Hacker News guidelines: https://news.ycombinator.com/newsguidelines.html
- Show HN rules: https://news.ycombinator.com/showhn.html
- LinkedIn User Agreement (effective Nov 3, 2025), section 8.2: https://www.linkedin.com/legal/user-agreement
- LinkedIn Professional Community Policies: https://www.linkedin.com/legal/professional-community-policies
- Stack Overflow generative AI policy: https://stackoverflow.com/help/gen-ai-policy
- Bluesky community guidelines (last updated Sept 19, 2025): https://bsky.social/about/support/community-guidelines
- Product Hunt commenting guidelines: https://help.producthunt.com/en/articles/10030102-commenting-guidelines
- DEV guidelines for AI-assisted articles (last updated April 8, 2024): https://dev.to/guidelines-for-ai-assisted-articles-on-dev
- FTC Endorsement Guides Q&A: https://www.ftc.gov/business-guidance/resources/ftcs-endorsement-guides-what-people-are-asking
- FTC Consumer Reviews and Testimonials Rule Q&A (16 CFR Part 465): https://www.ftc.gov/business-guidance/resources/consumer-reviews-testimonials-rule-questions-answers
Rules change. If you're reading this months later, open the links before you rely on any row of the table.