When you delete tweets by word on X, the result depends on where the filter looks and on what else you tell it to leave alone. Get either one wrong and the run removes posts you meant to keep, or leaves the ones you meant to remove. Both are mapped below, so the count you approve is the count you meant.
Circleboom's Keyword / Username or Hashtag filter checks every post on your X account for a word in its text, its @mentions, or its #hashtags, and deletes only the matches you approve. It connects to X under Circleboom's Enterprise developer status with X and reaches either your latest 3,200 posts or your full archive.
Below: where the match happens, which posts it can reach, and how to keep the ones you want.
Why a word filter on X catches more, or less, than you expect
A word filter matches characters, not meaning. It has no idea which of your posts used the word in the sense you now regret.
That gap is where most surprises come from. A filter that finds a former employer's name also finds the post where you congratulated a friend who still works there. A filter that finds a campaign hashtag may miss the posts where you wrote the campaign name without the hash sign.
The filter is exact about where it looks and blind about why the word is there.
Knowing where it looks is the part you can control. Once you know which fields are searched and which posts are in scope, you can predict the count before you see it. Then the review step becomes a check instead of a surprise.
The reasons people start a cleanup vary, from a job change to a privacy reset, and why people delete their old tweets goes into that wider question. The mechanics below apply to every one of those reasons.
Already know the word? You can find and delete tweets that contain a word in Circleboom and review the matches before anything is removed.
What does Circleboom's keyword filter search on X?
It searches three parts of each post: the tweet text, the @mentions, and the #hashtags. The same box accepts a plain keyword, an @username, or a hashtag, which is why Circleboom labels it Keyword / Username or Hashtag.
The filter exists on both of Circleboom's deletion routes:
- Delete My Last Tweets, the recent route, which pulls your newest 3,200 posts over the live connection with nothing to upload.
- Delete All Tweets, the archive route, which takes an uploaded tweet.js file and reaches back to your first post.
On both routes, every post starts selected. The keyword filter narrows that selection, and nothing is deleted until you confirm.
That design is the practical difference from doing the job on X itself, where every post needs its own menu and its own delete, and no total exists to check against. With Circleboom you can bulk-remove X posts by keyword and see the full count before the first one goes.
The connection matters too. Circleboom sends each deletion through the access X grants to its approved Enterprise developers, so the run follows X's own pacing.
The documented pace is 150 deletions per 15-minute window, or 14,400 a day, according to the processing times on Circleboom Twitter.
Does the filter catch hashtags and @mentions too?
Yes. Hashtags and @mentions are two of the three fields the filter searches, so a tagged account or a campaign tag is in reach, not only words in the sentence.
That matters most for two kinds of cleanup:
- A person or company you tagged often, where typing their @username finds the posts that mention them.
- A topic that lived in hashtags, such as an event, a cause, or a campaign, where the hashtag carries the word more often than the text does.
Election seasons are a common example of the second kind. Political hashtags pile up for weeks and then read very differently a year later, which is the situation behind deleting election tweets at once.
Two things are not spelled out: whether a plain word also matches inside a hashtag, and whether capital letters matter. Treat the word, the @username, and the hashtag as separate forms and check each count, which the next section covers.
Is the keyword match case-sensitive or whole-word?
Assume nothing either way. Whether the filter ignores capital letters, or catches a word inside a longer word, is something to test with a count before you confirm.
Both questions change the result. A partial match on a short word can pull in posts that contain it as a fragment of something longer. A case-sensitive match can miss the posts where you capitalized the word at the start of a sentence.
I check both with a comparison rather than a guess:
- Search X for your own posts with the word and note roughly how many come back.
- Run the same word in Circleboom and read the count it reports.
- Open a handful of the matches and confirm each one contains the word in the form you meant.
On the archive route, the Approve & Delete panel splits its count into tweets, retweets, and replies. On the recent route, the matches arrive as table rows, and each row's link icon opens that post on X for a direct read.
A count far above your search means the word is catching fragments. A count far below it means a form of the word is missing.
For the comparison search, the steps to search your Twitter history are worth reading before you start.
Which posts can a word run reach on X?
A word run reaches the post types you tick and the time window your route covers. On the archive route, all three post types start ticked: Tweets, ReTweets, and Replies. Untick one and the word run skips it.
The time window depends on the route. Delete My Last Tweets reads from your live account, and X's API stops at the newest 3,200 posts.
Anything older needs your archive: request it through X's archive download instructions and allow 24 to 48 hours before it is ready.
When the target posts date back years, you may also want to find old tweets on your profile first, to see how far back the word runs.
Only tweet.js goes into the upload, and the rest of the archive stays on your computer. Circleboom will not process a file that belongs to a different account than the one you connected.
Planning to wipe everything instead of one word? The Twitter archive eraser handles a full reset.
Can you delete the word but keep your best-performing posts?
Yes, by setting a maximum like or retweet count so higher-performing posts fall outside the run. The engagement filters work as Min and Max ranges, and the values are inclusive.
On the archive route, the like and retweet filters also show the highest count in your uploaded file. That number tells you the scale you are working with before you pick a limit.
An illustrative example shows how the range works. Say the highest like count in your file is 400, and most posts with the word have fewer than 20 likes. Set the like Max to 30, and every matching post with 31 or more likes stays on your profile while the rest are deleted.
Which engagement number to cap on the recent route
The recent route adds more ranges: reply, quote, bookmark, and impression counts, alongside likes and retweets. Each one protects a different kind of post:
- Likes: posts people approved of in the moment.
- Retweets and quotes: posts that traveled beyond your own followers.
- Bookmarks: posts people saved to come back to, often reference material.
- Impressions: posts that reached a large audience, liked or not.
Pick the metric that matches why you would keep a post. A how-to thread with modest likes but many bookmarks is the kind of post a like cap alone would delete.
Engagement caps keep strong posts out of a word run. They cannot tell you which weak posts still matter to you.
What if the word is in another language or script?
Type the word exactly as it appears in the posts, then use the Language filter to narrow or widen the run. The Language dropdown sits on the same filter step as the keyword box and defaults to All.
Two details change how this works. Language comes from the label X assigned to each post, not from Circleboom reading the text. And UND means Undefined, the label X gives posts without text, so a keyword run has nothing to match in those posts.
If a name or brand appears in two scripts, such as Latin and Cyrillic letters, typing one spelling gives the filter no reason to find the other. Run each script form as its own pass and compare the counts.
The Language filter also works in the other direction. When a word appears in English posts you want gone and in posts in another language you want to keep, set the language before you confirm.
If the language, not the word, is what you want gone, delete tweets by language handles that with the keyword box left empty.
When one post mixes two languages
X gives each post one language label, even when the post switches languages partway through. A post written mostly in Spanish with an English brand name in it still gets a single label, not two.
So a run limited to English can skip a post that plainly contains your English word. To catch those, run the word once with the Language filter on All and once with your language set.
The posts that appear only in the first count are the mixed ones, and you decide on those by hand.
Language-only cleanups on screen: how the Language dropdown narrows a bulk deletion when a word is not the right handle.
https://www.youtube.com/watch?v=ngb7q_w_95Y
How to delete tweets by word on X with the keyword filter
To delete tweets by word on X with Circleboom, pick the route that covers the posts and enter one form of the word in the keyword filter. Add the filters that protect what should stay, and confirm only after the count matches your search.
These four steps apply on both routes.
- Choose Delete My Last Tweets when every match is among your newest 3,200 posts, or Delete All Tweets with a tweet.js upload when some are older.
- Enter one form of the word in the Keyword / Username or Hashtag filter.
- Untick the post types that should stay, add a like or retweet maximum, and set a language if the word crosses languages.
- Compare the count with your own search, then click Delete my Tweets, or Delete Selected on the recent route.
The order puts every protective filter ahead of the confirmation, so the count you read in the last step already reflects what you chose to keep. A mismatch at that point is cheap to fix, and a deleted post is not.
Which X cleanup fits your word?
Find your case among the branches below.
If the word sits in your recent posts, use Delete My Last Tweets and read the table before deleting. If it goes back years, request your X archive first and use the tweet.js upload.
If the word lives mostly in hashtags or tags, run the #hashtag and the @username as separate passes. If it appears in more than one language or script, run each form and set the Language filter.
If your best posts contain the word, cap likes or retweets, plus bookmarks on the recent route, before you confirm. If every post from a certain period should go regardless of wording, a keyword is the wrong handle. It is faster to bulk delete tweets by date or type.
Every branch ends at the same checkpoint: a number you read while every post is still there to save.
→ Remove every tweet that contains your word
Delete tweets by word: X FAQ
How do I keep a few posts that contain the word?
Exclude them before you confirm. The archive route's Protect List holds any post you place on it out of the run, whatever it contains. On the recent route, Exclude Some Tweets or unticked rows plus Delete Selected do the same job.
Does deleting a post remove every trace of it online?
No. Deleting removes your post from X, as X's page on deleting posts describes, but replies, quote posts, and retweets from other accounts can remain.
For the wider cleanup beyond X, see delete your digital footprint.
Will a word run delete my likes?
No. The keyword filter works on your posts, meaning tweets, retweets, and replies, not on posts you liked. Circleboom handles likes with a separate delete-likes tool.
