The result
My 10 biggest measured original posts generated 2,067,969 impressions, 11,052 replies, and 10,417 likes. Nine of the 10 invited the reader to participate. Nine were text-only. The exception was a payout milestone where the images were the proof.
I did not pick these from memory or sort a recent analytics screen. I queried the full posting history stored for my account in ClimbX: 926 original posts from May 2 through August 26, 2026. Of those, 754 had a captured impressions snapshot and could be ranked.
The ranking uses each post's latest stored metrics as of August 27. Replies, self-replies, and hidden posts are excluded. That makes this my complete measured history in the database, not a claim about posts whose metrics were never captured.
The 10 biggest posts
I ranked these by impressions, then looked at what each post asked the reader to do. The lesson beside each post is the part I would reuse; the exact wording and subject should stay attached to the moment that made them work.
Use a familiar map, then leave one obvious space blank.
USA has ChatGPT USA has Grok USA has Claude USA has Gemini China has DeepSeek China has Qwen China has Kimi China has MiniMax Europe has?
The post makes a complicated topic, the global AI race, instantly scannable. Two complete lists establish the pattern, then a three-word question lets every reader supply the missing answer. It is topical, international, and easy to disagree with.
Name the audience and put the answer set on the screen.
Founders, which is the best place to buy a domain? - GoDaddy - Hostinger - Cloudflare - Namecheap - Others
“Founders” qualifies the reader immediately. The options are recognizable tools that people already have opinions about, and “Others” prevents the list from feeling closed. Nobody has to invent a reply from scratch; they only have to pick a side.
Make the reader the protagonist, not the tool.
Tell me one thing you can do that CLAUDE cannot do yet
The hook uses a famous AI product, but the real subject is the reader. It invites people to defend their own value in one sentence. That is more personal than another prediction about what AI will replace, and the one-answer constraint keeps the reply easy.
A specific product gap beats a broad industry question.
Why doesn’t Claude have an image or video generation model yet?
It names one product and one conspicuous missing capability, a narrower target than a broad “where is AI going” question. Users can answer from experience, speculate about strategy, or argue with the premise. Specificity gives the conversation somewhere to start.
Put the promise beside the lived experience.
genuine question: AI was supposed to make us work less so why does it feel like everyone is working 10x more now?
The tension is recognizable before the reader reaches the question: AI promised less work, but ambition expanded to fill the saved time. It works because it describes a feeling many builders already have and asks them to explain it.
A tight constraint can turn a reply prompt into a game.
I am a founder scare me with 1 word
The identity is clear, the emotion is universal, and the rule is one word. Readers can join in seconds and compare their answer against the hundreds already sitting in the replies. A one-word limit turns a reply prompt into a fast, scannable game.
When the number is the story, show the receipt.
First X payout: $828.77 🤯🤯 Took me 81 days to get here from never using X before.
This is the only top-10 post with images, and they carry the argument: the screenshots verify the exact payout named in the text. Amount, elapsed time, and starting point make the milestone concrete. Proof turns a celebration post into a credible one.
Give creativity a vivid target.
I need a cool startup name that sounds like it just raised $100M.🤔
“Suggest a startup name” is generic. “Sounds like it just raised $100M” gives the prompt personality, status, and a mental image. People know what kind of answer belongs, but still have room to be original.
Ask for one useful answer, not an entire playbook.
Founders who cracked distribution: how? Just the thing that worked.
The follow-up line removes stories and theory: give me the thing that worked. It produced fewer replies than most posts in this list, but 236 bookmarks, the second-highest count in the top 10, suggest readers valued the answers as a resource worth saving.
Turn self-promotion into a challenge people want to attempt.
sell me your startup in two words max.
Founders already want to share their product. The two-word limit makes that familiar behavior entertaining for everyone else too. It creates a public compression test: can the value proposition survive without the pitch deck?
What the winners have in common
- 1Participation is built into the idea. Nine posts ask for an answer, choice, suggestion, or pitch, and the ask is built into the sentence structure itself, not tacked on as a closing “what do you think?”
- 2The prompts have edges. One word. Two words. One thing. Pick from named domain registrars. Constraints reduce the cost of replying and give the response section a recognizable shape.
- 3The topics sit inside my audience's daily life. AI models, coding tools, domains, founder fear, startup names, distribution, and X payouts are all things builders already argue about before I ever post on them.
- 4Proof images and decorative images are different formats. Nine winners needed no media. The payout post needed its screenshots because they verified the claim. An image earns its place when removing it would remove evidence.
- 5One viral post can distort the whole picture. The Europe-and-AI post produced 65.7% of all top-10 impressions. I should learn from its contrast and open loop, but treating 1.36 million impressions as the new baseline would be a strategy built on one exceptional event.
The playbook I would reuse
Start with a subject your audience is already comparing, buying, fearing, or trying to understand. Find one unresolved gap or tradeoff inside it. Then make the response cheap: provide named choices or ask for one constrained answer. If the post makes a personal claim with a real number, attach the receipt. Otherwise, let the text stand alone.
Familiar subject + specific tension + low-friction response
Add proof only when proof is part of the story, and reuse the decision architecture that made people stop and participate rather than the exact wording of a single winning post.
What I would not conclude from this list
A vague question does not earn a reply on its own; it still has to give the reader something specific to answer. The payout post's screenshots did real work as proof, and readers rewarded that with 1,230 likes. Reply count only measures conversation: the distribution post pulled fewer replies than the playful prompts, but its 236 bookmarks, the second-highest in the top 10, say people saved it to use later. Reach, conversation, and utility measure different things.
The useful conclusion is narrower: my audience responds when a post is specific enough to recognize, open enough to join, and simple enough to answer immediately. That is the pattern worth repeating, one clear ask per post.
Find the patterns hiding in your own posts.
ClimbX connects your post history, performance data, and current outliers so its learning loop can show what is actually working for your account, and use those patterns when drafting the next post.
Read next
- Do AI images hurt your reach on X? I tested it on my last 100 posts. - I ran a small experiment: more images on my posts, many of them AI art. The data from 100 posts: AI illustrations and other decorative images cut reach by 48% on generic posts and 65% on one-liners - while real proof images (payout screenshots, charts, records) gained 78%. The full numbers, the clean split, and the ClimbX feature this experiment killed.
- 81 days on X, 6,900 followers, and a first payout of $828.77. - The exact playbook behind 81 days of grinding X: 480 posts, 23,300 replies, 6M impressions, and a first monetized payout of $828.77. Plus why that payout was the least valuable thing it produced.
- Outliers as training data: how ClimbX learns what works at your size. - Why we draft from posts that broke out for accounts 2 to 5x your size, how the cohort data refreshes, and how the learning loop tightens with every draft you ship - all framed by what the X algorithm actually rewards.