How to Search Twitter by Date With since: and until:

Short Answer: Search the Next Day Too

To search Twitter/X by date, use `since:YYYY-MM-DD until:YYYY-MM-DD`. Treat it as a half-open range: `since:` is inclusive, and `until:` is exclusive.

For one exact day, put the day you want in `since:` and the following calendar day in `until:`. `since:2024-05-13 until:2024-05-14` means posts on or after May 13 and before May 14, not two full days.

If the post's displayed date could be shifted by timezone, screenshot time, or where you viewed it, search the previous and next day too. For a remembered date of 2024-05-13, the copy-ready 3-day check is `since:2024-05-12 until:2024-05-15`.

Use this page for the date boundary itself. If the account is the stronger clue, use the account-specific old-tweets guide. If wider date windows still miss a post you know should exist, use the search-misses guide.

X's advanced search fields support date ranges, words, hashtags, and accounts. Community operator references such as Advanced Search on Twitter are useful for copy-ready syntax, but they are not a guarantee that X search exposes every matching public post.

The useful habit is boring: keep the account or phrase stable, widen the dates first, and only then add media, hashtag, or link filters.

Example Query: The Result Still Needed a Record

A stored field check used this input through CarryFeed's public search endpoint: `from:OpenAI GPT-4o since:2024-05-13 until:2024-05-14`. The observed response returned public rows that included OpenAI posts about GPT-4o access and live demos.

One returned source was OpenAI's public GPT-4o access post. The count is not the proof, because X search can filter and reorder results. The useful evidence is smaller: input, observed result, and limitation.

That is the difference between finding something and having evidence. Do not cite the result count as coverage. Keep the public URL, query, observed author, visible date, capture date, and the caveat that the result list can change.

When One Day Returns Nothing

A one-day search can fail even when the post is public. The remembered date may come from a local timezone, a screenshot, a news article, a reply timestamp, or the day you saw the post rather than the day it was published.

X also documents search rules and restrictions, so search results should be treated as filtered behavior rather than a complete public archive.

When the first query is empty, do not add five new operators. Keep the strongest clue and move the date boundary. Search the previous day, the next day, or a 3-day window before changing the phrase. If that still fails, loosen the phrase. If it still fails, remove media or link filters. If the direct source opens as protected, deleted, suspended, restricted, or login-only, stop and look for another public citation.

Widen Dates Before You Add Words

The temptation after an empty result is to add more memory: a longer quote, a hashtag, `filter:media`, a domain, maybe a second phrase. That usually makes the search harder to understand.

A cleaner pass keeps one anchor stable and widens only the dates. If the account is strong, keep `from:handle`. If the phrase is strong, keep two distinctive words. Let the date window tell you whether the memory was too narrow.

1. First pass: `from:handle phrase since:YYYY-MM-DD until:next-day`; `until:` is the first date outside the range.
2. Timezone pass: keep the same account or phrase and search previous day through day after next.
3. Second pass: use a 3 to 7 day window with the same account or phrase.
4. Third pass: use the month, then add one extra clue at a time.
5. After a useful result appears, open the public source before changing the query.
6. If the source matters, save the query and capture date beside the URL.

If You Know the Account, Start There

A public account is usually a stronger clue than another keyword. If you know who posted, start with `from:handle` and a date range. Add topic words after the date window is close.

This is where searches often get worse. People remember an account, then add a long phrase, hashtag, media filter, and one-day range at the same time. When that returns nothing, they do not know which constraint failed.

If the job is broader than date search and you mainly know the account, use the companion guide How to search a user's old tweets by account, date, and clue. This article is specifically about making the date boundary behave.

Add Hashtags, Media, and Links After You Have Candidates

Hashtags, media filters, and link/domain clues are valuable, but they are risky first filters. They can remove the right post if the attachment type, hashtag spelling, short link, or quoted domain is not indexed the way you remember.

Use them after the account/date or phrase/date query produces a manageable set of candidates. That keeps the date window from becoming tangled with an attachment guess.

For media-heavy posts, the date query gets you to the candidate. The source URL decides the next step: open the viewer for context, use an image or video downloader only when public media is exposed, or export the result row for review.

Save the Dated Result Before You Keep Searching

The query explains how you found the post, but the public URL is the part you can reopen. Save it before you keep experimenting with wider ranges.

For a date search, keep the query, source URL, visible author, visible date, capture date, date window used, and whether you had to search a previous or next day.

If the found source is going into AI notes or a research brief, pair this habit with Twitter Metadata for AI Agents. Date search only helps if the final public source remains checkable after it leaves the search page.

Where CarryFeed Fits

CarryFeed should not pretend that public X/Twitter search is complete. Its useful job here is narrower: make date fields easier to compose, open public results when they appear, and keep the final URL beside the date range that found it.

Use the advanced search page when you want date fields instead of hand-typing every operator. If the account is the real clue, switch to the old-tweets guide. If a known public post still refuses to appear, switch to the search-misses guide.

In the example check above, the useful output was not just a list of posts. It was a dated record the reader could revisit: query, result URL, author handle, visible text, capture date, and the warning that result order can change.

Stop at the Public Boundary

This guide is for public posts. X's page on public and protected posts draws the boundary: protected posts and protected media are limited to approved followers.

Do not use date search as a way to imply hidden access. If the source asks for an authorized follower, shows unavailable, belongs to a protected account, or only survives as a screenshot, record the limit and stop.

The honest outcomes are enough: found public source, wider date window needed, alternate public citation needed, or unavailable public source.