The thread that sends readers back to your blog doesn't summarise the post — it takes the single finding that surprised you while writing, proves that one claim inside the thread with the actual number, and leaves the implementation behind the link. The distinction matters more than it sounds: when Buffer turned 10 blog posts into threads and also posted each one as a plain link tweet, the threads won 63% more impressions and 54% more engagements — but only 8% more link clicks, a result skewed by outliers, with most individual threads getting fewer clicks than their own link tweet. The thread format buys you reach. It does not automatically buy you readers.
That gap is the whole problem. You can run the "repurpose your post into a thread" playbook perfectly, watch the impressions climb, and still send fewer people to the post than a single link tweet would have. So it's worth being precise about what actually causes the click, and what is folklore.
Buffer's test is the cleanest public version of the exact thing you're doing every week: ten blog posts, each published as a thread and as a link tweet a few days apart, results recorded shortly after. Threads won on reach and interaction by wide margins. On clicks they essentially tied, and Buffer named its own suspect: "For the Twitter threads, the link was attached to the last tweet or in the middle. Our followers don't see the link immediately and have to scroll through the whole thread before seeing the link."
Two caveats, because your readers will find them if you quote this: the sample is ten posts from December 2018, on an algorithm since replaced more than once — Buffer itself warned "the results might not be replicable all the time" — and the click number was noisy enough that Buffer called it inconclusive.
What survives is the shape of the finding, not the exact percentages: reach and referral are separate outcomes, and a thread optimises the first one by default. Accept that and the design question changes. You stop asking "how do I make a good thread?" and start asking "what has to be true for someone who has just read eight of my tweets to still want the ninth thing — the link?"
For years the standard answer was mechanical: X throttles posts with links, so bury the URL in a reply. That advice is now officially dead. On 28 July 2026, then-head of product Nikita Bier told Mark Zuckerberg on-platform, "Hello Mark, you do not need to put the links in replies anymore." Hours later, asked by Paul Graham whether X still penalises posts with links, Elon Musk replied: "We haven't for over a year." No formal announcement accompanied either statement, and creator reaction ran to scepticism — several noted that Bier's "anymore" quietly confirmed the penalty had existed, and said they would keep testing placement themselves.
The code backs the leadership up, in a way that's less reassuring than it first looks. The Phoenix ranking model xAI open-sourced on 20 January 2026 predicts a set of engagement types — favourite, reply, repost, quote, click, profile click, dwell, follow, and the negative signals — and scores posts as the weighted sum of those probabilities. There is no URL rule anywhere in it. But there is also no prediction for an external link click: the generic click signal can't tell on-platform navigation from a trip to your blog. A link post that sends someone away generates a weaker signal than one that keeps them tapping — so suppression gets learned rather than written down.
Bier described the same mechanism in plain language on 19 October 2025, announcing an in-app browser test: "a common complaint is that posts with links tend to get lower reach. This is because the web browser covers the post and people forget to Like or Reply, so X doesn't get a clear signal whether the content is any good."
The independent numbers, meanwhile, stayed grim. Buffer's analysis of 18.8 million posts from 71,000 accounts published through its own tool between August 2024 and August 2025 found link posts from non-Premium accounts sitting at a 0% median engagement rate since March 2025, while Premium accounts still registered around 0.28% on links and 0.90% on plain text. Both things can be true at once: no rule in the code, no penalty per the leadership, and near-zero measured engagement for free accounts. Which leaves you with no switch to flip and no trick to game — the link post has to earn its engagement.
One more thing that didn't change: the algorithm weights everyone still quotes. If you've seen "Retweets ×20, Replies ×13.5, Link Clicks ×11" in a 2026 thread, those trace to the 2023 twitter/the-algorithm release — whose README publishes no numerical weights at all — and describe a heavy-ranker architecture since retired. xAI's current repo says the Grok-based transformer ranks with "no manual feature engineering for content relevance," and keeps the weight values undisclosed. Optimising against those numbers means optimising against a system that no longer exists.
Solo founders routinely benchmark themselves against outcomes that were never available to them. Across that same 18.8-million-post dataset, median impressions per post were under 100 for regular accounts, about 600 for Premium and over 1,550 for Premium+ — and regular accounts had a median engagement rate of 0%, meaning at least half got no visible interaction on a typical post. For comparison, Ahrefs — a company with a corporate following that promotes every post with paid ads — reports 113 link clicks per day across its tweets, not per post. Zoom out further and Chartbeat data across roughly 1,350 publishers showed Twitter's share of referrals falling from 1.9% in April 2018 to 1.2% in April 2023, with small publishers losing 98% of their Twitter page views.
None of that is an argument for skipping X; it's an argument for aiming at the right target. A good thread from a small account converts a handful of genuinely interested people into readers of one post, and the value compounds in your profile and your replies rather than in a traffic spike. Twenty of the right people beat two thousand impressions that bounce.
Here's the structural problem with the thread that lists your ten takeaways: it closes every loop it opens. The reader already has the value; clicking through would be re-reading. You've optimised for what X rewards — dwell, replies, bookmarks — at the direct expense of what you wanted.
There's a subtler failure on the other side, and it's the one "curiosity gap" advice walks you into. A 2025 Scientific Reports analysis of the Upworthy Research Archive — 8,977 A/B experiments covering 35,910 headline instances — found the relationship between concreteness and clicks is curvilinear, not linear. Where baseline headlines were vague, adding specificity raised clickthrough by roughly 5.5%; where they were already concrete, adding more detail cut clickthrough by roughly 9.9%. The authors' summary: "headlines that convey just the right amount of information maximize clickthrough rates at scale."
Read those two failure modes together and the design rule falls out. Vagueness loses. Completeness loses. What wins is a thread that is fully concrete about one thing and silent about the rest.
X's own economics now point the same way. Original Content Rewards, announced on 8 August 2026 to replace Revenue Sharing, states that "if your content regularly incorporates material created by others, you'll need to contribute meaningful original value," and excludes reposted material lacking meaningful transformation. The copy-paste summary is exactly the format the platform is now designed to under-reward — the same "unoriginal at volume" concern behind Google's scaled content abuse policy, arriving on a social platform.
So what do you actually write? Four moves, in order.
1. Pick the finding that surprised you, not the thesis. Berger and Milkman's 2012 study of what gets shared found that high-arousal emotion predicts virality while surprise, interest and practical utility are each independently linked to sharing. Your post's thesis is rarely surprising — it's the position you already held. The surprise is the fact that made you rewrite a paragraph. That's the thread.
2. Prove that one claim inside the thread — with the number and its scope. "Threads don't get clicks" is a vibe. "Buffer tested 10 posts both ways; threads won 63% more impressions and 8% more clicks, and most threads individually lost on clicks" is a claim someone can check, argue with, and quote. Naming the sample size isn't a weakness — on X it's what makes technical readers trust the rest of the thread, and it inoculates you against the reply-guy who looks it up.
3. Stop when the claim is proved. This is the discipline the Upworthy curve demands. Once your reader believes the finding, every additional detail is subtracted from their reason to click. The method — how you'd apply it, the checklist, the caveats, the second and third findings — stays in the post.
4. Give a reason to click now. The folk theory here is the "open loop" — leave it unfinished and they'll come back. Treat that gently: a 2025 meta-analysis of the Zeigarnik and Ovsiankina effects found no reliable memory advantage for unfinished tasks, though the tendency to resume an interrupted one did hold up. So don't bank on being remembered this evening. Make the last post state exactly what's behind the link and who it's for.
There's a fifth move that isn't about writing: check that the numbers you're quoting still resolve. A dead source or a figure that moved is far more expensive on X than in a blog post, because the correction arrives in your replies. It's one reason our own agent won't save a draft until every cited URL resolves and ships a verification report with the post — the claims you pull into a thread should have been checked before the post went live, not after someone quote-tweets you.
The writing sets up the click; the mechanics decide whether it happens.
It's a fair question in 2026, when the platform keeps building reasons to stay. X Articles opened to all Premium subscribers on 7 January 2026, long-form posts run to about 25,000 characters, and Exclusive Threads let creators monetise directly on-platform. If distribution is the only goal, publishing natively is genuinely competitive.
The counterweight is what happens after the timeline moves on. Otterly's analysis of over a million citations across ChatGPT, Perplexity and Google AI Overviews in January–February 2026 found community platforms taking 52.5% of citations against 47.5% for brand domains, with blog and personal sites at 3.5–10% depending on the engine — small, but real, and x.com doesn't appear among that dataset's top-cited domains. Your thread is a moment; the post is the artifact that gets indexed, quoted and reused. That's why structuring posts to be cited by answer engines is worth more than the referral count suggests, and why a blog still earns its keep in a zero-click world.
How long should the thread be? There's no credible public dataset behind the "5–7 posts" numbers that circulate — every version traces back to assertion, not measurement. Use the claim as your ruler instead: as many posts as it takes to prove the one finding, and no more. If a post isn't evidence for that claim, cut it.
Should I still put the link in the first reply? X's leadership says it's unnecessary as of late July 2026, and there's no URL rule in the open-sourced ranking code. But the change was never formally announced, so treat it as a hypothesis about your account rather than a settled fact — alternate placements over a few weeks and read your own clicks-per-impression.
Does a thread beat a plain link post? On reach, usually — that's what Buffer's ten-post test measured. On clicks, not reliably. If your goal for a given post is readers rather than followers, a strong caption with a link is a legitimate choice, and Buffer's data shows link posts still do meaningfully better for Premium accounts than free ones.
What if the post has no surprising finding? Then the thread will struggle, and that's useful information about the post. A piece with no claim worth defending in public usually summarised what was already ranking — the fix belongs upstream, in topic selection.
Can I reuse the same finding in a second thread? Yes, with a different frame — Ahrefs stages two formats per post on separate days for exactly this reason. What doesn't work is three threads making the same claim the same way, which is where a record of what you've already published earns its keep.
One post, one finding, one thread, one number to check afterwards — then write the next post out of whatever the replies argued about. That loop is unremarkable in any single week and decisive over a year, which is also its weak point: it only compounds if it actually happens every week, and the part that slips first is the post, not the thread.
That's the gap Magic Share is built for. The agent wakes at 06:00 UTC, researches and drafts what your queue has due, verifies every cited link before the draft is saved, and holds it for your approval — then shares the approved post's thread to X. Your first three posts are free, and nothing publishes without you.
Magic Share researches, writes and fact-checks posts like this for any site — point it at your URL and review your first draft today.
Plant your first post