Shares on X
Copy link is the strongest positive weight in X's code. Here's how to earn it.
In one sentence: in X's published For You weights, a predicted "copy link" share is worth 20, about 30× a like (0.66). Shares are the strongest positive signals in the file.
"Who needs this?"
That's the whole approach: one short question at the end of a post that's genuinely useful. It doesn't beg. It just makes readers think of one specific person, and then they DM the post or copy the link, the two highest share weights in X's code (copy link 20, DM 5; like 0.66).
Use it when: the post is a checklist, template, tool, how-to, or a warning that saves someone time or money.
Don't use it when: the post is an opinion, a joke, news, or about yourself. A question after a post nobody needs feels like bait and can earn "not interested" (−52.272) instead.
1. The three share types vs like and repost
- Signal (param.rs, 2026)Weight
- Share via copy link+20.0
- Share via DM+5.0
- Share (share menu)+2.0
- Repost+1.0
- Favorite (like)+0.66
Exact lines: ShareViaCopyLinkWeight, ShareViaDmWeight, ShareWeight, RetweetWeight, FavoriteWeight. Ratios: copy link ÷ like = 20 ÷ 0.66 ≈ 30.3; DM ÷ like ≈ 7.6; share ÷ like ≈ 3.0.
Important: these weights multiply the predicted probability that a viewer takes the action, not the raw number of shares. X's comment in param.rs: the weights "reflect a combination of how much an action is valued in ranking and typical propensities of these actions". So rare actions get bigger weights partly because they're rare. 30× doesn't mean one copy-link counts as 30 likes.
2. What each one means
- Share via copy link (20)
- Someone copies the post's URL, usually to paste it somewhere else: a group chat, Slack, a newsletter, another app.
- Share via DM (5)
- Someone sends the post to a person inside X's direct messages.
- Share (2)
- The share action itself, e.g. opening the share menu. The code names it separately from DM and copy link. The exact UI event it maps to isn't documented uncertain.
- Repost (1) and like (0.66)
- Public, one-tap actions inside X.
3. Why shares weigh so much
- From the code: weights combine how much X values an action with how often it happens. Shares are listed under "Engagement" in the README's prediction list.
- Interpretation (not stated in the code) uncertain: a copy or a DM is a private, deliberate act. Someone thought "a specific person needs this." That's a strong sign of value that's hard to fake with one tap.
4. How to write posts people copy and forward
Write for a specific second person
People share when the post solves a problem for someone they know. End with a question that brings that person to mind:
- "Who needs this?"
- "Send this to the friend who still …"
- "Save this for your next …"
Tag-a-friend vs better alternatives
Why: "tag a friend" is public and feels like bait. The better version points to a private share (DM or copy link, the two highest share weights) and gives a reason. Overly baity posts also risk "not interested" (−52.272) and mutes (−58.8). See Negative signals.
Saveable formats
- Checklists: "7 checks before you post a thread"
- Templates: fill-in-the-blank hooks or replies
- Tools and links: a free tool, with what it does in one line
- Comparisons: X vs Y in a short table or list
- Step-by-step: numbered, so it's easy to follow later
5. Examples (illustrative)
These are writing examples I made up to show the format. They're not real posts, and they come with no results.
Example (AI): "5 prompts I use every morning to plan my day with Grok. Copy them. Send this to the friend who still plans in their head."
Example (Tesla): "Before your first Tesla road trip, check these 6 settings. Who needs this before their next trip?"
Example (growing on X): "A 4-line reply template that gets read. Save it, then send it to someone who replies 'great post' all day."
6. Mistakes
- Asking for shares with nothing worth sharing.
- "Tag 10 people" spam, which invites mutes and reports.
- Hiding the value behind a link only. The post itself should be the thing worth forwarding.
- Walls of text nobody can skim later.
- Reading 30× as "one share = 30 likes". It's about predicted probabilities.
Open Share Hooks: 18 question variants
7. FAQ
- Can I see how often my posts are shared?
- X's analytics show some share metrics. This page only covers what the code says.
- Are these weights fixed?
- They're defaults declared in param.rs. The README says experiments and configuration can change values uncertain.
- Were shares weighted in 2023?
- Shares aren't in the April 5, 2023 heavy ranker list in the recap README. They show up as separate weights in the 2026 code.
- Does a link in my post hurt?
- Open link has weight +0.2 in param.rs. No link penalty appears there uncertain.