Instagram impressions vs views: the 2025 change

There is a column in a great many social media reports labelled “impressions” that has not contained an impressions figure since April 2025.

The metric was not renamed, deprioritised or moved. It was removed from the API, on a published schedule, with a replacement that counts something different. If your reporting spans that date and nobody marked the break, your trend lines have a discontinuity in them that looks exactly like performance.

The dates, from Meta’s changelog

The sequence is documented and specific:

• 21 January 2025. Meta announces that impressions will be removed, starting with version 22.0 and above.
• 21 April 2025. Full deprecation across all versions. After this date, requests for media created on or after 2 July 2024 return an error. Requests for older media continue to return data in v21.0 and earlier.

Source: the Instagram Platform changelog. The industry mirrored it at the time in near-identical language (Sprinklr, Emplifi, Iconosquare).

Note the awkward middle state in that second bullet. From 21 April 2025 there were accounts whose older posts still returned impressions and whose newer posts errored, from the same request, depending only on when the media was created. The changelog says versions 21.0 and earlier keep returning impressions for media created on or before 1 July 2024, and as of September 2026 it gives no retirement date for those versions. Tools handled that inconsistently, which is one reason archives from 2025 vary so much between platforms.

Views is not the same measurement

Views is the count of times a post, story or Reel was played or displayed, repeat views by the same account included. It replaced impressions for feed posts and stories and plays for Reels on 21 April 2025, and it is reported only for accounts you manage.

This is the part that gets skipped, and it is the part that matters.

Before April 2025 there were three separate counts:

After April 2025 there is one:

Three definitions were collapsed into one. A Reel that starts playing as somebody scrolls past and a carousel that renders in feed now both increment the same counter, where previously they incremented different ones with different thresholds.

So the honest statement about a views figure and an impressions figure from 2024 is that they are not comparable. Views is generally the larger number, but not by a stable ratio, because how much larger depends entirely on your format mix. An account that shifted toward Reels across the change will show a step up that is partly real and partly definitional, and no amount of staring at the chart will separate the two.

What this breaks

Year-on-year comparisons. Any chart crossing April 2025 with a single continuous line is comparing two different measurements. The step is an artefact of the change and will be read as growth or decline by whoever sees it.

Anything divided by impressions. Engagement rate by impressions was a common formula. It cannot be calculated for recent posts, because the denominator does not exist. Tools that still offer it are computing it over views and, in most cases, not saying so.

Historical benchmarks. A benchmark published before 2025 that quotes an impressions-denominated rate is not a target you can measure yourself against any more. The measurement it was built on is gone.

Stored archives. If you warehoused impressions data before the change, it is still valid for what it was. It is not a series that views can be appended to.

What to do about it

Mark the break. Put a visible annotation at April 2025 on any chart that spans it. This is the single highest-value fix and it takes minutes. A footnote does not do the job, because the person who misreads the chart is the person who did not read the footnote.

Move to reach for continuity. Reach was not affected. It counted unique accounts before the change and it counts unique accounts now. If you need one distribution metric that runs cleanly from 2023 to today, reach is the only candidate, and it is the better metric anyway for most questions people actually ask. It is still available only for accounts you manage, as it always was.

Restate the denominator on every rate. If your engagement rate used to be over impressions and is now over views, it changed definition in April 2025 even though the label did not. Say so once, in the report, where the number is.

Audit any tool still showing impressions. For post-April-2025 media, that column is historical data, views relabelled, or an estimate. All three are defensible. Not knowing which one you are looking at is not.

Why Meta did this

Meta’s changelog records the change and not the reason. The effect is simplification, and the simplification reading is ours. Instagram had accumulated three overlapping ways of counting roughly the same thing, split across formats for historical reasons, and the boundaries had stopped making sense. A Reel plays; a carousel displays; a video needed three seconds. Merging them into views is defensible and matches how the app now presents performance to creators.

The cost lands entirely on anyone maintaining a time series across the boundary, which is to say every agency reporting to a client with a year-on-year slide. Meta absorbed none of that cost. It was distributed to everyone downstream, and mostly absorbed by not being noticed.

The short version

Impressions was removed on 21 April 2025 and replaced by views, which merges three previously distinct counts. It is not a rename and the two are not comparable. Reach is unchanged and is the metric to trend across the boundary.

Any impressions figure for recent media is historical, relabelled or estimated, and a report that does not say which has made a claim it cannot support.