How creators responded to Instagram's 5-hashtag limit

What changed

Instagram limited posts and Reels to 5 hashtags in December 2025, saying that fewer, more targeted hashtags can improve performance. Reported coverage also describes reduced distribution for posts above the cap; Instagram has not published that mechanism, so this page does not treat it as established.

The adoption curve

Baseline of 59.9% over the limit across the 6 months before the announcement, against 5.2% since the change landed in Feb 2026. That month is derived from the series rather than chosen: it is the first month under a fifth of baseline.

Is this just different accounts?

This is the question that should be asked of any before-and-after, and it is the one that has invalidated findings like this before. Posts from before the cap are largely back-filled history from accounts that connected later, so the two windows do not contain the same accounts, and the pooled change could in principle be composition rather than behaviour.

Restricted to the 99 accounts with at least 3 posts on each side, mean hashtags per caption go from 6.33 to 3.31, and the share over the limit from 48.3% to 6.5%. Same accounts, same change.

The control also disagrees with the pooled figures on a second measure, which is the clearest illustration of why it is here. The share of captions carrying no hashtag at all falls in the pooled data and rises among the same accounts, while the monthly series shows no trend either way. We report that as a direction disagreement rather than a number, because the pooled version is composition.

What we cannot see

• Who is in the sample. Accounts connected to Reelyze, most of them the creator's own rather than competitors someone chose to watch. A describable group, not a random sample of Instagram - and one that is MOVING. Between 4 and 7 September 2026 the companion follower-multiple sample went from 222 accounts to 218 while the share that are creators' own rose from 81.5% to 84.9%, because people untrack competitors and nobody untracks their own account. The composition below is therefore dated like every other figure here: on a long enough run this stops being a sample of tracked accounts at all. This supports the claim that the creators we can observe changed their behaviour, not that Instagram creators did.
• Captions only. Counts come from captions. Hashtags placed in the first comment are not visible to us, so every count here is a floor rather than a total. Instagram has not published whether hashtags in the first comment count toward the 5, and the sources claiming they do and claiming they do not are both guessing.
• Truncation, checked. 97 of 16,752 captions end in an ellipsis, so provider-side clipping is not what makes these counts a floor. The first comment is.

Three things we cannot explain

Published as open questions. A dataset that answers everything it raises is a dataset that has stopped looking.

1. November 2025 was already falling before the announcement

   Over-limit captions went from 58.6% in October to 48.3% in November, and exactly-five rose from 7.3% to 12%, three weeks before the 19 December announcement.

   Consistent with a quiet test rollout. There is no evidence for that, and on 333 posts it may be noise.

2. The over-limit share is not flat, and it has risen

   1.4% in February and 1.2% in May, then 8.3% in July and 6.6% in August on 2,622 posts. A large rise on a large sample, not a steady residual.

   Loosened enforcement, a client that does not enforce the cap, scheduled posts written before it, or a shift in which accounts we sample. All unknown.

3. Around a quarter of captions carry no hashtag at all, and the two ways of measuring it disagree in DIRECTION

   Pooled, the no-hashtag share falls from 28.7% to 23.6% across the announcement. Within the same accounts it does the opposite and rises slightly, 28.8% to 31.5%. Month by month it simply bounces between 19.5% and 28.1% with no trend.

   The pooled fall is composition, not behaviour: it is the same confound the control exists to remove, showing up on a second measure. Whatever the level means, it is not an effect of the cap. Consistent with tags living in the first comment, which we cannot see.

Method

16,752 posts read, 16,722 counted, 30 excluded for having no caption. Hashtags are counted distinct and case-insensitive, because Instagram treats #Fitness and #fitness as one tag and counting both would overstate how many of the 5 slots a caption spends. Measured 2026-09-07. The month containing the measurement date is excluded as partial.

Reproduce it with scripts/derive_hashtag_cap.py --as-of YYYY-MM-DD.

These figures cannot be reproduced later, and we would rather say so than let you find out. The posts we read are deleted when a creator untracks an account, so the source shrinks over time and a re-run returns different numbers for reasons that are not a re-measurement. On 2026-09-07 it fell by roughly a fifth in a single day. Treat every figure here as a dated reading of a sample that no longer exists in the same form, rather than a query you can repeat. These figures change continuously as tracked accounts are added and removed: treat them as a reading at the moment of derivation, not a stable measurement of that day. The same-account control below moved by five accounts within two hours of being derived. An earlier version of this page described the query as reproducible. That was wrong, and this replaces it.

Related: do hashtags still work on Instagram, how to pick your five, and the free hashtag generator.