Most AI social media tools measure what is easy to count. Across 2,209 Reels the number that actually tracked reach was engagement rate
Engagement rate tracks reach far better than view count does. Across 2,209 Reels, 0.39 against 1.10 views per follower. Measured 24 September 2026.
What the numbers say
There are a great many AI tools for social media and most of them report the same things, because those things are the easiest to fetch: follower count, view count, posting frequency. Across 2,209 Reels we sorted by engagement per thousand views, which is likes plus comments divided by views, and the reach difference is large. Reels under 20 engagements per thousand views reached a median of 0.39 times their follower count. Reels at 70 or more reached 1.10. That is about 2.8 times, on the same measure, in the same corpus. Comments move with it: the median Reel in the lowest band carried 0.4 comments per thousand views and in the highest band 1.3. Raw view count, the number almost every dashboard puts largest, mostly reports how big the account already was.
Read engagement rate before you read views. It is the number that moved with reach in this corpus, and unlike view count it is not mostly a restatement of your follower count.
Sample of 2,209, measured on 2026-09-24. Scrptly production database, queried 2026-09-24. 2,209 Reels deduplicated by shortcode, each with over 1,000 views plus a like count, a comment count and its account's follower count.
The measurement
Engagement per thousand views is likes plus comments, divided by views, times a thousand. Views per follower is the Reel's views divided by its own account's follower count. Reels under 1,000 views are excluded because a rate over a tiny denominator is noise. The four bands hold all 2,209: 572 plus 723 plus 623 plus 291. One caution before reading it: this is an association, and the honest reading is that a Reel people want to engage with is also a Reel the feed carries further. It is not a recipe where collecting likes causes reach.
| Engagement per 1000 views | Reels measured | Median views per follower | Top tenth reaches | Median comments per 1000 views |
|---|---|---|---|---|
| Under 20 | 572 | 0.39 | 14.26 | 0.4 |
| 20 to 39 | 723 | 0.40 | 4.82 | 0.8 |
| 40 to 69 | 623 | 0.60 | 8.19 | 1.1 |
| 70 or more | 291 | 1.10 | 16.68 | 1.3 |
One Reel, beat by beat
This is what the top engagement band looks like from the inside. One real Reel from the corpus, not a composite: 1,932,527 followers, 1,181,198 views, 272,395 likes and 1,498 comments, which is 231.9 engagements per thousand views. Its reach was 0.61 times the following, so it is not a breakout by reach at all. It is a two person conversation built entirely around one question the viewer will have an opinion about, and the timeline below runs to 39 seconds.
| At | What happens |
|---|---|
| 0:00 to 0:03 | One speaker poses a binary question. Nothing else. The viewer is now picking a side |
| 0:04 to 0:07 | The second speaker commits to one side and starts justifying it |
| 0:08 to 0:15 | Build by stating the common belief, the one the viewer probably holds |
| 0:16 to 0:30 | Contrast it with rapid everyday examples. This is the longest stretch and the persuasive core |
| 0:31 to 0:39 | Restate the thesis as the payoff. No call to action, because the disagreement is the call to action |
What to do next
- Open your own Insights and work out engagement per thousand views for your last twenty Reels: likes plus comments, divided by views, times a thousand. Then find your band in the table above.
- If most of your Reels sit under 20, the problem is not distribution. Something in the content is not giving people a reason to react, and no tool that reports view counts back at you will surface that.
- Treat comments as the harder and better signal. The median Reel in the top band carried 1.3 comments per thousand views against 0.4 in the bottom band, and comments take more from a viewer than a like does.
- Be careful what you conclude. This is an association measured across Reels, not an experiment. Chasing engagement with comment bait will move the numerator without changing whether the Reel was worth watching.
- One honest limit: these are Reels from accounts somebody chose to analyse, so the corpus skews towards content that already worked.
- Scrptly is an AI tool for social media and it deliberately does not report your follower count back to you. It reads Reels that performed and returns the hook score, the transcript and the beat order. The first analysis does not need an account.