Skip to content
ReplyMan logoReplyMan
Log inSign upGo to dashboard

Blog

How to Reply on X Without Getting Flagged as Spam

By Sayan De · · 12 min read

Blue cover reading: Replies that stay visible on X, not flagged as spam. The rules that keep replies out of Show probable spam

Last checked October 16, 2026, against X's Developer Guidelines and the sources listed at the end. X publishes very little about how it ranks or hides replies, so each claim below says whether it comes from X, from our own tests, or from third parties.

If you reply to strangers on X to find customers, your replies can end up in one of three places. They show up. They get folded under "Show probable spam", where almost nobody reads them. Or your account is limited or suspended. X documents the third outcome. It says almost nothing about the second.

This guide covers what X has published, what we measured ourselves, and the rules we use for replies drafted in ReplyMan. The first rule matters most: a person should read and post every reply. It is written for founders and personal-brand accounts who reply to strangers by hand or with a drafting tool.

Key Takeaways

  • X's published rules ban browser automation and unsolicited automated replies. AI-written replies need X's approval before they are deployed. A person reading and posting each reply keeps you outside those rules.
  • Spam signals that third-party guides agree on are one-line replies, repeated text, fast bursts and promotion. X has not confirmed which ones it uses.
  • We tested two busy threads: Premium replies ranked above nearly every non-Premium reply. A free account's reply starts low, whatever it says.
  • A search ban and a hidden reply are different problems. Checking one does not tell you about the other.
  • There is no official safe number of replies per day. Start low, pace the replies out, and raise the number only while replies keep showing in threads.

What X can do to a reply

People use "shadowban" for three different things. They have different causes and different fixes.

Outcome What happens Does X tell you? How sure we are
Suspension or lock The account cannot post Yes, on the account Documented by X
Search ban Your posts do not appear when someone searches from:yourname No Third-party guides describe it, and we could reproduce it
Hidden reply Your reply sits behind "Show more replies" or "Show probable spam", or at the bottom No Visible in the product. Why a given reply is hidden is not published

The ranking code X open-sourced in 2026 confirms that predicted negative reactions lower a post's score. It lists "not interested", "mute author", "block author" and "report". It also describes a spam filter that applies to posts from accounts the viewer does not follow (xai-org/x-algorithm). That describes the feed, not a reply-by-reply explanation. Treat any guide that gives exact thresholds with suspicion.

What X has actually published

What the guidelines say

These are quotes from X's Developer Guidelines, read on October 15, 2026. They govern apps and automated accounts. A person writing and posting a reply is not what they describe, which is why rule 1 below exists.

  • Browser automation: "Non-API automation (scraping, browser automation) results in permanent suspension."
  • AI replies: an app that generates and posts replies "requires prior approval from X before deployment", and "deploying AI-generated replies without approval is a violation, even if the content itself is helpful."
  • Automated accounts: "Don't: Send unsolicited DMs, replies, or @mentions." Automated accounts may reply only "if user engaged first. Max 1 reply per interaction."

How X has enforced it

  • February 23, 2026: X restricted replies through its API on the Free, Basic, Pro and Pay-Per-Use tiers. An app can reply only if the original author mentioned or quoted the account (X Developers, PiunikaWeb). Enterprise and Public Utility apps are exempt.
  • July 24, 2026: X's then head of product said it had removed 42,000 accounts that automated replies with chatbots, adding that "AI engaging with users without a human in the loop runs counter to X's mission" (summary at TrustAudience). X published no detection method.
  • May 2026: X's help page limits unverified accounts to 50 posts and 200 replies a day, according to MediaNama. That is a ceiling for every kind of reply. It is not a target, and X says the limits may drop under heavy load.

X's rules also treat repeated, duplicated and unsolicited replies to many accounts as spam (X Rules and best practices).

The rules

In short: you post it, it says something specific, it is never repeated, it goes out slowly, once per person, with no hashtags, and the product comes last. The numbers below are our own starting points, not X's. X publishes no safe daily count. We built them from X's published rules, the sources at the end, and the tests in the next section.

1. A person reads and posts every reply

Do not use a script, a scheduler that replies for you, or an extension that clicks Post. (How ReplyMan handles this on X, and what it does with your accounts.) Everything in X's guidelines above is about automation. Opening x.com with your reply filled in and pressing Post yourself keeps a human in the loop. It is also the only approach X has said it will not suspend for.

2. Answer the post's point in your own words

One-word and one-line replies ("GM", "Great point!", "This") are what Sorsa's guide lists generic or low-effort replies as a trigger for reply deboosting. It is a vendor's claim, but it costs nothing to follow. Say something about this post that could not be pasted under another one.

An illustration of the difference (an example we made up, not data), for a post asking how people find their first customers:

Likely folded Likely read
"Great point! We built a tool for this, check it out." "Answer a few founder questions a day with one concrete tip each, and after a few weeks people start asking what you use. If you built a tool for it, say so then, and say it is yours."

The second reply is specific to the question, and a product mention, when it comes, is last and disclosed.

3. Never send the same reply twice

Duplicate and near-duplicate replies to different people are the clearest case in X's spam rules. Change the substance, not just the opening. In ReplyMan's drafting rules, a reply "is written for this post only: no stock opening, sign-off or pitch you would reuse elsewhere."

4. Pace replies by how old and established the account is

OpenTweet's guide advises against "excessive liking, retweeting, or commenting in short bursts", and short gaps and big bursts are the pattern X's automation rules are written against. New accounts have the least reputation to lose from, so go slowest there. These are our defaults:

Account Replies a day Gap between replies
New: under 30 days old, or under 50 followers 5 15 minutes
Growing: in between 10 8 minutes
Established: over 180 days old and 500+ followers 20 4 minutes

Vary the gaps. Replies that go out exactly every eight minutes are a rhythm. If replies start disappearing, drop a tier before you do anything else.

5. One reply per thread, one per person per week

A second unsolicited reply to the same person reads as pestering, and X's guidelines say automated accounts get one reply per interaction. Wait until they answer you before replying again.

6. No hashtags and no extra @mentions

Reply to the person you are replying to, and nobody else. Uninvited @mentions are on X's list of prohibited behavior for automated accounts, and hashtags in a reply add nothing a reader needs.

7. Mention your product last, and say it is yours

Answer the question fully first. If your product fits, mention it at the end and say you built it or work on it. Reply with the product line removed: it should still help. Keep links out of replies while the account is new. Claims about links conflict: one analysis of X's published ranking weights finds link clicks count in a post's favor, while other guides report heavy reach penalties. We could not confirm either, so we treat links as a risk for new accounts and leave it there.

8. Do not open a new account to keep going

If an account is restricted, a second account that replies in the same places is ban evasion, and both can be suspended.

What we measured: Premium replies rank first

On October 15, 2026, we pulled the replies to two busy posts from one large account through the twitterapis.com API (/tweet/replies, sorted by relevance, X's default). "Premium" means the API's is_blue_verified flag, which is the blue check. One thread had 79 replies and the other 89, after removing the ads the list mixes in: they show up as entries that are not replies to the post.

  • Thread 1: all 53 Premium replies ranked above all 26 non-Premium replies.
  • Thread 2: the first 69 replies held 67 Premium and 2 non-Premium accounts, and the last 20 replies were all non-Premium.

Separately, three accounts in the second thread returned no results for a from: search, which is how a search ban shows up. Their replies still ranked 9th, 22nd and 28th of 89.

Two points follow from this. First, a free account's reply on a busy thread starts at the bottom, whatever it says. Pick threads where a reply can be seen, such as posts with tens of replies, not hundreds. That is our judgement, not a tested result. Second, a search ban and a hidden reply are separate. Checking one tells you nothing about the other.

The limits of this test: it covers two threads under one author, and it used a third-party API's ordering, which may differ from what a logged-in viewer sees. It shows a pattern. It does not show X's rule. You can see the same thing yourself: open a busy thread logged out and look at who sits at the top.

How to check whether your account is restricted

  1. Search ban: log out or open a private window, search from:yourname, and open the Latest tab. If you have posted recently and nothing appears, you probably have a search ban (Sorsa's guide).
  2. Hidden replies: reply to a post from an account you do not follow, then view that thread logged out or from a second account. If your reply is missing or behind "Show more replies", it is being hidden. A second account makes this much easier.
  3. Account status: open Settings, Your account, Account status. It lists labelled restrictions with an appeal button. Silent restrictions do not appear there, so a clean page rules out little.
  4. Checker tools: twitterapi.io tests the search ban only and says it does not check reply deboosting. Sorsa tests the search ban and search suggestions, and says reply deboosting needs a manual test. No checker we found detects hidden replies.

If your account is restricted

Third-party guides agree on the recovery steps. X has not confirmed them.

  1. Stop replying for two to three days. More activity can restart the penalty.
  2. Delete promotional or repeated replies from the last 48 to 72 hours.
  3. Complete the profile: photo, bio, header, verified email and phone.
  4. Under Security and account access, remove third-party apps you do not use.
  5. Appeal through Account status if there is a label.
  6. Come back slowly, with replies that add something and no links for a while.

Guides report that a search ban lasts about two to seven days and reply hiding about three to fourteen, with repeated triggers lasting weeks (Sorsa). These are vendor claims, not X's.

What ReplyMan does on X

  • You press Post. When you approve a draft, x.com opens with the reply filled in. ReplyMan does not post to X for you.
  • Replies come in small batches. Posting reminders email you a few replies that are safe to post now, at the times you pick, so they never go out in a burst.
  • The drafts follow rules 2 to 7: no one-liners, no repeated reply, no hashtags, no extra @mentions, the product mentioned last with disclosure, and one reply per thread or person.
  • It enforces the pacing in rule 4. ReplyMan applies the account's tier limits and gaps from the table above, holds a reply until it is safe to post, and shows why on the card. It also limits how often you mention your product or post the same link.
  • It shows your 90/10 split. Your account row carries how your last 30 days divided between product mentions and helpful replies — green under 10%, amber at or over — so you can see when the balance tips too far toward promotion.
  • It checks the account when you connect it, and refuses a handle that does not exist or is suspended.
  • It reads the account once a day and finds your replies after you post them, even when X search hides them. A folded reply is marked, and posting pauses if X suspends the account or hides several replies.

See how X works in ReplyMan.

Frequently asked questions

Does X officially confirm shadowbans?

X does not document "shadowbans" as such. It does show a "Show probable spam" section in threads, which collapses replies it rates as lower quality (TrustAudience), but it does not say why a given reply is in it. Labelled restrictions appear under Account status. Silent ones do not.

Does an X Premium subscription protect my replies?

It lifted replies in our test, but nothing in X's guidelines exempts Premium accounts, and a blue check does not rule out a bot (TrustAudience).

Can I post replies through the X API?

Only if the author mentioned or quoted your account, since February 23, 2026. Strangers have not, so API replies to them are rejected. AI-generated replies also need X's prior approval.

How many replies a day is safe?

X publishes no safe number. Its 200-a-day ceiling for unverified accounts is for all replies. Start at the low numbers in rule 4 and move up only while your replies keep showing in threads.

Sources

Sayan De builds ReplyMan, a reply tool that finds public posts on X, Reddit and other networks and drafts replies for you to approve. This guide describes ReplyMan's own X rules, so weigh it accordingly. The measurements are ours, from October 15, 2026, and the sources above are linked so you can check them. More about ReplyMan, or contact the author.