Do links hurt your reach on X?
Updated
Probably not by a fixed penalty anymore. X's owner and its then head of product both said in July 2026 that link posts aren't penalized, and we found no link penalty in the published ranking code. Link posts can still reach fewer people if readers respond to them less. There's no official X document that settles this, so test it on your own account.
What X's leaders have said
On July 28, 2026, Nikita Bier, then X's head of product, told Mark Zuckerberg he didn't need to put links in replies anymore. Paul Graham asked whether X had stopped penalizing them, and Elon Musk answered, "We haven't for over a year." PPC Land has the full exchange, and it notes this wasn't a formal announcement.
X's messaging hasn't been fully consistent. In October 2025, Bier denied that links were penalized at all. "Anymore" and "we haven't for over a year" suggest something did change at some point. Bier also left the role in August 2026, so this reflects what X said then.
What the published code shows
We searched X's published code on November 3, 2026, at commit b220448.
- Opening a link has a positive weight. The parameters file sets it at 0.2, which is small next to 5 for a reply.
- There's no fixed URL penalty in the ranking code we read. The scoring formula is a sum of predicted actions, and no link multiplier appears in it.
- The system does know about links. A
has_urlfield is set for posts with an outside link, and it's passed along in the data sent to the service that reorders results. The field's name doesn't say how it's used, and it was the only such field we found.
A trained model can also learn patterns that nobody wrote down as a rule. So "no written penalty" doesn't prove "no effect." If people tend to like or reply less to link posts, the model can predict that and rank those posts lower.
Why link posts can still get less reach
This part is our reasoning, not something X has measured for you.
A post that sends people away gives them less reason to reply in the thread. Replies carry a far higher weight than link opens, so a bare headline and a URL gives the model little to work with. Bier has made a related point: people who open a link often forget to like or reply to the post, so the system gets a weaker signal.
Be careful with precise claims. Some sites publish figures like a 90% reach penalty for links without any source. We couldn't trace those to X's code or documents.
How to post links now
Treat the post as the content and the link as an extra.
Here's an invented example:
- Weak: "New post on pricing: [link]"
- Better: "We raised our price from $9 to $19 and lost 4 of 212 customers. Here's what we told them and the email we sent: [link]"
The better version is useful even if nobody clicks it. That gives readers something to respond to on X.
A two-week test for your account
Your audience matters more than any general rule, so run this yourself:
- Pick one kind of post, and write 10 posts of that kind with a link you'd want to share.
- Alternate them. Put the link in the post for one, and in your first reply for the next.
- Post at similar times, and leave a few days between posts on the same topic.
- After a week, record impressions and replies for each, plus link clicks if your analytics show them.
- Compare the medians, not the single best post.
Ten posts won't prove much on a small account, because one post can swing the result. Treat a clear gap as a hint and keep testing. Clicks may matter more to you than reach anyway, since a post with fewer views and more visits to your site can still be the better one. Our engagement rate calculator makes the comparison quick.
For the bigger picture, read how the X algorithm works.