Why retweets no longer show up in a Twitter search
X moved its search API to a new index and keyword searches stopped returning retweets. What a report keeps, what it loses and which numbers move.

Pull a hashtag report today and compare it with one from last year. The retweets are missing, and the count of people who took part is lower. Nothing broke.
In May, X moved the search endpoints of its API to a new index. The migration brought real improvements, faster queries, far fewer errors and new operators for precision. It also brought one change that nobody reporting on X can ignore: a search by keyword, hashtag or phrase no longer returns retweets.
This is not one tool deciding something. It applies to every product built on the X search API, because the API simply stopped sending them.
Tell us what you need to measure and we will tell you what a report shows today, not what it used to show.
Tell us what you need to measure
What a report keeps
The retweet count on every tweet. That number still arrives with each original post, so the total for a campaign is calculated exactly as before. If forty thousand people retweeted your hashtag, the report still says so.
Real impressions are untouched, and this is the part worth understanding. A retweet does not carry impressions of its own. It feeds the impressions of the tweet it amplifies. So the reach a conversation actually had is measured today the same way it was measured two years ago, because it never depended on seeing the retweets as separate items.
Likes, replies, quotes and the original tweets themselves all arrive as they always did.
What a report loses
Retweets as individual rows. They were in the export as their own records, and they are not any more.
The list of who retweeted. This is the real loss. You can no longer see which accounts amplified a tweet, because the API does not send them, so there is nothing to build that list from.
If your work depended on that list, on mapping who spreads a message rather than who writes it, that analysis is not available from keyword search any more.
What moves in the numbers
Two figures come out lower than they used to, and both for the same reason.
Potential reach and potential impressions drop. Those were always built by adding up the followers of everyone who posted or retweeted. Without the retweeting accounts there are fewer follower counts to add, so the estimate falls.
Contributor counts drop too. Somebody who only retweeted used to appear as a participant in the conversation. Now they do not appear at all.
Neither of those is a loss of accuracy. Potential reach was an estimate built on the assumption that every follower sees everything, which they do not. The number that describes what happened, the impressions the platform itself reports, is the one that did not move.
Comparing across the change
A series that crosses May compares two different things, and pretending otherwise produces a drop that somebody will read as a bad quarter.
Say so in the report. One line explaining that retweets left the dataset in May is enough to stop the conversation going the wrong way.
Then compare what is comparable. Original tweets, impressions, likes, replies and the retweet counts all behave the same on both sides of the line. Contributors and potential reach do not.
Reposts or retweets
X renamed them reposts. Everyone still says retweets, the API field is still called that, and both words mean the same thing here. Nothing about the naming changes what is or is not in a report.
What we do about it
We pass on exactly what the API gives us, and we say which figures the change affects rather than quietly showing a smaller number.
Every report still ships with the export behind it, gathered within the Twitter compliance rules, so any figure can be traced to the tweets that produced it. See the plans or tell us what you need to measure.