Use Grok Bot to Find 100+ Posts and Send 30+ Replies Daily

Input seed accounts & ICP → Grok Bot expands 2nd/3rd-degree accounts → human replies for 2 days → bot learns style → daily morning/evening reply candidate reports

Who

Small-account operators on 𝕏 looking to grow through high-quality replies

You get

100+ candidate posts daily, 30+ human replies sent; bot only searches and suggests

You need

X Pro

Steps

1. Define Your ICP

   Before starting, precisely define who you want to attract. If your ICP is vague, Grok Bot will return random accounts. The author references an ICP exercise (7 questions about who your avatar is and what they want) but does not include the full text; you will need to prepare this yourself.

2. Pick 5 Seed Accounts and Build a 1st-Degree List

   Choose 5 accounts you already engage with (author examples: Trader XO, Follis, EllioTrades, Koroush AK, Luckshury). Manually review who replies to them and who they reply to. Add accounts closest to your ICP to a Private List in 𝕏 Pro. This list becomes the core data source for Grok Bot.

3. Tell the Bot What Good and Bad Accounts Look Like

   Before letting the bot expand your list, set clear filtering boundaries or results will be poor. The criteria are formalized in the next step's Prompt 1: good accounts need keyword in bio, recent activity, follower count and verification, and human-like engagement; bad accounts include signal sellers, flex accounts, engagement farms, brand/exchange promos, and extreme political or hate speech.

4. Let Grok Bot Expand 2nd/3rd-Degree Accounts

   Fill your seed accounts, 1st-degree list, ICP, and topic into the prompt below and send it to Grok Bot. The bot will traverse who interacts with your 1st-degree accounts (2nd degree), then repeat for those new accounts (3rd degree), outputting a ranked table. You can then ask the bot to add results to your 𝕏 Pro list.

   Execution prompt · Source text。Grok Bot chat input; replace [@handle1...], [keyword], [my topic] placeholders before sending

   Copy prompt

   My seed accounts: [@handle1, @handle2, @handle3, @handle4, @handle5]
   My 1st degree list: [paste the handles you added manually]
   
   My ICP: [one line, e.g. systematic traders and quants who are great at trading but underbuilt as a brand]
   My topic: [e.g. quant trading / hedge funds / AI]
   
   GOOD (must match most of these):
   - [keyword] in bio
   - Talked about [my topic] in at least 5 of their last 10 posts
   - 1.5K+ followers AND verified
   - Original post in the last 3 days
   - Replies to other people, and the replies use real details and sound human
   
   BAD (drop no matter what):
   - Signal sellers / Telegram funnels
   - Flex accounts
   - Engagement farms
   - Brand or exchange promo accounts
   - Not ultra political or hate speech
   
   STEP 1: 2ND DEGREE
   For each account on my 1st degree list, go through who replies to them and who they reply to. Find 5 accounts that match GOOD and none of BAD.
   
   STEP 2: 3RD DEGREE
   Repeat Step 1 for every account you found. Stop there. Never list the same account twice.
   
   OUTPUT
   One table:
   handle | followers | degree | found via | verified | last post date | topic fit (/10) | why they fit (one line)
   
   Sort by topic fit. Under the table, describe my ICP in one paragraph based on what the matches have in common.
   
   Research only. Do not follow, like, reply to or DM anyone.

5. Reply Manually for 2 Days to Create Training Data

   Before the bot can assist with suggestions, it needs to learn your habits. For 2 days, browse your 𝕏 Pro list and reply naturally to posts. Don't overthink it—the goal is authentic data. Note why you replied to or skipped an account (to show expertise, personality, or a contrarian take). These notes help the bot build guardrails later.

6. Have the Bot Study Your Replies and Build Guardrails

   Starting day 3, send the prompt below to Grok Bot to analyze your last 2 days of replies. The bot outputs: what account/post types and reply styles work, which accounts to add/move/remove, and ranking rules for your daily report. Run this every morning and iterate the guardrails continuously.

   Execution prompt · Source text。Grok Bot chat input; replace [2] with actual number of days before sending

   Copy prompt

   Study my replies from the last [2] days.
   
   Map out:
   - The types of accounts I keep replying to (role, topic, follower range)
   - The types of accounts I struggle to reply to or skip
   - The types of posts I reply to (question, opinion, lesson, chart, story, news, meme)
   - How I naturally reply (length, tone, questions, trading examples, jokes)
   - Which replies got the most impressions, likes and replies back
   - Which types of people engage back when I reply
   
   OUTPUT
   - 3 bullets on what's working
   - Accounts to add, move up, move down or remove from my list, with a reason
   - Ranking rules for my daily report: which posts to show first and which to cut
   
   Research only. Do not post, reply, like, follow or DM anyone.

7. Get Daily Morning/Evening Reply Candidate Reports

   From day 3 onward, use the prompt below morning and evening to scan your 𝕏 Pro list and generate a report of posts worth replying to. Reports rank by freshness first, then topic fit, with a skipped section. Spend one focused 10-15 minute window per report reading, writing, and manually sending replies. Reply to anyone who replies to you.

   Execution prompt · Source text。Grok Bot chat input; replace [7am], [5pm], [list name] with actual values before sending

   Copy prompt

   Every morning at [7am] and evening at [5pm], check my X Pro list [list name] for new original posts since the last report.
   
   For each post:
   - Link + handle
   - Why it's worth replying to (one line)
   - 2-3 reply angles, written the way I naturally reply (use my guardrails)
   - Extra context if it's linked to a news event, or to something they or others posted in the last day or week
   
   Rank by freshness first, then topic fit.
   Add a short "skipped" section at the bottom showing what you filtered out and why.
   Never repeat a post from an earlier report.
   
   Drafts only. Do not post, reply, like, follow or DM anyone.

No prompt needed — just follow the steps.