Engagement Farming on X: What It Is, How to Spot It, and Why It Breaks KOL Campaigns

Engagement farming is inflating likes, replies, and reposts with tactics that have nothing to do with real interest: engagement pods, reply circles (rings), bots, giveaway bait, and rage bait. On X, it makes an account look more influential than it is. So before you pay a creator for reach, look at who engages with their posts and how often the same accounts show up.

X has been fighting this for a year. In July 2026, it started removing creators from revenue sharing if they ask for engagement three or more times, and in April it said it was suspending 208 bots a minute. The farming still works well enough to fool buyers. For this article, we pulled the replies under eight recent posts from a crypto account with about 12K followers. 36 accounts replied under at least four of the eight, and one replied under every single post.

If you buy creator posts in crypto, you'll meet all seven patterns below, often on the same account.

What are the 7 types of engagement farming on X?

Most farmed accounts mix several of these. The first three fake the audience. The other four bait real people into engaging with content they don't care about.

1. Engagement pods and Telegram raid groups

An engagement pod is a group of accounts that agree to like and reply to each other's posts. Researchers at NYU and Drexel studied pods on Instagram and found 1.8 million post links shared in more than 400 pods, with 111,455 users taking part. Most of those pods were organized on Telegram.

Crypto runs the same thing as "raids". Someone drops a post link in a Telegram group with a time window and two or three talking points, and members are expected to like, reply, and repost before the window closes. Raid bots track who completed the task and rank members on a leaderboard. A raid can be honest community support for a project's own post. It turns into farming when a creator's paid post is raided to look more popular than it is.

2. Reply circles

A reply circle is a smaller, permanent version of a pod: the same 20 to 50 accounts reply under each other's posts every day. You see it as familiar names under every post, usually within the first hour. X's rules call this "reciprocal inflation" and name "follow trains", "decks" and "Repost for Repost" as examples.

3. Bot likes nd bought engagement

Fake engagement is cheap. In a 2025 experiment, NATO's StratCom centre bought 156,000 fake views on X for €10. Four weeks later, 43% of the engagement it bought on X was still live. AI made the bots harder to spot. A 2023 Indiana University study found a network of 1,140 ChatGPT-powered accounts promoting crypto sites that replied to and reposted each other, and a standard bot detector missed almost all of them.

4. Follow-back and giveaway bait

"Can I get it back?", "I'll follow everyone who replies", "Like and repost to win 100 USDT". These posts trade a small reward for an engagement, so the replies come from people who want the follow or the prize. Facebook started demoting this kind of post in 2017, with "LIKE this if you're an Aries!" as its example. X added a hard rule in July 2026: ask for engagement like this three or more times and you're out of the revenue-sharing program.

5. Rage bait

Rage bait is a post written to make people angry enough to reply. Oxford named it its 2025 Word of the Year after usage tripled in twelve months. The replies are real, but most come from people arguing with the creator. A creator who grows this way has a big, irritated audience, and your sponsored post lands in front of it.

6. "gm" farming

"GM CT" posts get dozens of "gm" replies, and replying "gm" to everyone is a cheap way to get noticed. On its own, it's harmless crypto culture. It becomes a problem when you read a reply count as interest. In our sample, though, the pure "gm" replies were a small share (about 5%). The farmed account's replies were mostly one-line comments about whatever project the post was shilling, and 11% of them repeated, word for word, a reply someone else had already posted.

7. InfoFi yap farming

InfoFi platforms such as Kaito's Yaps paid users in their points or project’s tokens for posting about different topics. Accounts learned to post whatever earned points. One loophole reportedly paid out for repeatedly commenting the word "reply". On 15 January 2026, X's then-head of product Nikita Bier said X would no longer allow apps that reward users for posting, blaming them for "a tremendous amount of AI slop [and] reply spam". Within hours, Kaito shut down Yaps and Cookie DAO wound down Snaps. The accounts that farmed those leaderboards are still on X, and so are the habits: long project threads written for points, and reply chains under each other's posts.

Engagement farming vs organic virality: how to tell them apart

A viral post and a farmed post can have the same like count. The difference shows up in who engages, what they write, and when. The numbers in the last column come from our comparison of two crypto accounts of similar size, described below.

One signal alone proves little. Big accounts have fans who reply to everything, and a small niche account can get more replies than likes after a good question. Look for three or more signals pointing the same way.

Is engagement farming against X's rules?

Most of it is. X's platform manipulation and spam policy opens with "You can't artificially inflate your own or others' followers or engagement." It specifically bans:

• trading or coordinating to exchange follows or engagement (pods, circles, follow trains);
• coordinated activity across multiple accounts to influence conversations;
• selling or buying followers, reposts, likes, mentions or poll votes;
• bulk, aggressive, unsolicited replies and mentions.

Rage bait and "gm" posts aren't banned. X handles them through ranking and payouts instead. Since July 2026, accounts that ask for engagement three or more times lose access to revenue sharing, and X says its Grok model now detects these posts. Earlier, X cut payouts to aggregator accounts and planned a permanent deduction for accounts that put "BREAKING" on every post (Social Media Today). The ranking model X open-sourced in January 2026 predicts negative reactions such as blocks, mutes and reports along with likes and replies, though it doesn't publish how much each one counts (Business Today).

The bot purges have been large. X said it removed 1.7 million reply-spam bots in one week of October 2025. In November 2025, it added "About this account", which shows the country an account is based in (Axios). For crypto, X's head of product at the time went further and said in April 2026 that "80% of crypto is simply bots". He gave no data for that figure, so treat it as frustration from someone who sees the problem daily.

Outside X, buying fake engagement can also break the law. Since 2024, the US Federal Trade Commission has banned selling or buying fake indicators of social media influence, such as bot followers or views, when the buyer knows they're fake and uses them commercially.

Why engagement farming breaks KOL campaigns

When you pay a creator, you're paying for their audience’s attention. Farmed engagement inflates the numbers you price on and adds nobody who could use your product.

Start with CPM, the price per 1,000 views. A post that costs $1,000 and gets 100,000 views has a $10 CPM. That's in line with what our own prediction-market campaign paid ($9.86). Now suppose part of those views came from bots and pod members. The price stays the same, but you've bought fewer real readers:

The formula is simple: real CPM = quoted CPM ÷ (1 − farmed share). Faking the views costs almost nothing by comparison. At the price NATO's researchers paid, 100,000 fake views on X cost about €6.

Replies are worse. Many teams pick creators by "engagement", because likes and replies look harder to fake than views. A group of 36 accounts replying to every post makes a small creator look like they have an active community. When you sponsor that creator, the same 36 accounts reply to your post too. You get a thread full of "this looks bullish" and no one opening your app.

The damage doesn't stop at the wasted fee:

• You learn the wrong lesson. The report shows good engagement, so you rebook the same creators next quarter. We wrote about this pattern in why KOL campaigns get views but no users.
• Your brand sits next to the farm. Real readers who open the replies see bots and copy-paste comments under your announcement.
• X may demote the post. X now ranks posts on predicted blocks, mutes, and reports, and it removes farming accounts in waves. A paid post on an account X is about to downrank can underperform its own history.

How to detect engagement farming on any X account

You can check a creator by hand. Use their last eight to ten original posts, skipping replies and reposts.

1. List who replies. Write down the handles under each post. If a handful of names appear under more than half of the posts, you've probably found a ring. Fans exist, but a real audience changes from post to post.
2. Read the replies. Look for generic praise, emoji-only replies, and lines that could sit under any post. Search a suspicious reply in quotes: if the same sentence appears under other accounts' posts, it was pasted.
3. Open five reply profiles. Check what they post when they're not replying, when the account was created, and the country in "About this account". A profile that only replies, joined recently, and sits in a country far from the creator's audience is a red flag.
4. Compare likes, replies, and views. Organic posts usually get several likes for every reply. Replies close to the like count, or dozens of replies per 1,000 views, point to arranged engagement.
5. Check the first hour. Raids happen in a time window. If most of the replies arrive within minutes from the same accounts, the post was probably dropped in a group.
6. Look at the creator's own bait. “Can I get it back," “follow everyone who replies,” and constant giveaways tell you how the numbers were built.
7. Compare sponsored and organic posts. If paid posts get as many replies as the creator's best organic posts every time, ask who is replying to the ads.

For the follower side, see our guide on how to spot fake influencers and the Twitter (X) audit. The full pre-payment process is in our influencer vetting checklist.

Doing this for one creator is fine. Doing it for a shortlist of 50 takes days. Oznaka (oznaka.ai) is a creator vetting tool for crypto teams: it scans an X account's recent posts and shows organic reach and organic replies alongside the raw numbers, plus niche match, a fair price, and an overall Fit Score. Scan a KOL free.

What a pod looks like in the data

We took two crypto accounts with similar audiences, about 12K and 14K followers, and collected the direct replies under each one's last eight original posts. Account A opens every post with "GM CT … Can I get it back" and then promotes a project. Account B is a researcher who posts his own analysis. We don't name either account.

We collected 866 replies from 508 accounts under Account A's eight posts. 36 of those accounts replied under at least four of the eight posts, and they wrote 18% of all replies. Under Account B's posts, we collected 46 replies, and no account replied under more than three of them. Account A's regulars also replied faster: a median of 99 minutes after the post, against 268 minutes for everyone else.

This is one comparison, not a study. Account A gets far more replies, which makes repeat names more likely, and our search caught about 86% of its replies and fewer of Account B's. The pattern is still the one to look for: a fixed core of accounts under every post, plus a crowd of one-off repliers around it.