Impressions vs Engagement on X: Which Number to Chase

2026-08-29 · 6 min read

X shows impressions publicly on every post, which has quietly made it the metric everybody optimises for and the metric that means the least. A post with 90,000 impressions and 12 likes did not do well. It was shown to a great many people who did not care.

What each number actually is

An impression is the post being rendered on a screen. It does not require the person to read it, stop scrolling, or even notice it. It is a delivery count.

Engagement is any deliberate action: a like, a reply, a repost, a profile click, a bookmark, an expansion of the post to read it in full. All of these require the viewer to decide to do something, which is why they carry information that impressions do not.

The ratio, and what is normal

Divide total engagements by impressions. Most accounts sit somewhere between 1 and 5 percent. Under 1 percent consistently means your posts are being distributed to people who do not want them, which is almost always a mismatch between the opening line and the substance.

Above 8 percent usually means one of two things: a small, unusually devoted audience, or a post that hit something specific. Both are good, and neither scales automatically.

Why chasing impressions makes an account worse

Impressions respond to provocation. A post designed to make people angry gets shown widely because arguing generates replies, and replies are the highest-weighted signal in the system. So an account optimising purely for impressions drifts, post by post, toward being annoying.

The cost arrives later and is invisible in the numbers you are watching. Blocks and mutes are weighted very heavily against you and they are cumulative, so an account that farms outrage for six months builds up a penalty it cannot see and cannot easily undo.

Bookmarks are the underrated one

A bookmark is the strongest voluntary signal a reader can send short of a reply: it means they intend to come back. It also correlates with the kind of post that gets found later through search, which is the only form of long-tail distribution X has.

If your bookmark count is high relative to likes, you are writing reference material, and reference material is worth repeating and expanding rather than treating as a one-off.

Where paid engagement belongs

The useful case is narrow and it is about the ratio, not the total. A post sitting at 4,000 impressions with two likes reads to a human as unpopular, and people genuinely do decide whether to engage partly on what other people did. Adding a proportionate number of twitter likes fixes the appearance of a post that is otherwise fine.

The failing case is buying impressions or views on a post nobody engages with. That makes the ratio worse — more delivery, the same tiny numerator — and the ratio is both what the algorithm reads and what a person notices. It is the one place where buying the bigger number actively works against you.

Whatever you add, keep it in proportion to the account. An account with 300 followers and a post at 40,000 views is not ambiguous to anyone reading it.