How often does a Twitter alert bot check for new tweets?
Quick answer: it depends entirely on the bot, because the checking interval is a design choice and not something X imposes — and in X-Relay, we re-check a tracked account as often as every 300 ms. That interval is called the polling cadence: the gap between one look at a tracked account's timeline and the next, and the thing that bounds how long a brand-new tweet can sit there before anybody downstream knows it exists. It is not the same as how quickly the alert reaches your chat, which also depends on X serving the post in the first place and on Discord or Telegram accepting the message. Accounts on a standard plan sit in the normal rotation; Turbo — a paid add-on, not a separate plan — moves them into the fast one.
The two numbers we publish
| Number | What it is | Where it comes from |
|---|---|---|
| Every 300 ms | How often we re-check a tracked account — polling cadence | A property of how the polling schedule is built, not a measurement |
| 281 ms | Detection as fast as 281 ms — our fastest measured detection | Best of 51,655 detections, 72 h on production to 29 Aug 2026 |
There is no third number on this page, and the omission is the point. The first answers the question in the title — how often we check. The second is a record from measurement, kept separate on purpose. Everything below explains what each one covers, what Turbo does to the first of them, and what we refuse to put a figure on at all.
Why how often a bot checks sets the ceiling
Any tracker that reads X on a schedule has a hard limit set by that schedule. If a system looks at an account once a minute, a post made one second after a check waits fifty-nine seconds before anyone downstream even knows it exists — and nothing built after that point recovers the lost minute. The interval is the ceiling.
That is why we state the cadence rather than an arrival time. We re-check a tracked account as often as every 300 ms, so the window in which a fresh post can sit unnoticed is small by construction, on every single check, for every account in that rotation. It is not an average taken across a good day. It does not improve when traffic is light or collapse when it is heavy. It is simply how often we look.
It is also the number that separates real trackers from the things people try first. An RSS bridge rebuilding a feed every fifteen minutes and a browser tab someone refreshes by hand are both, in this sense, the same product: the delay is the interval, and no amount of fast plumbing afterwards makes up for it.
Polling cadence is not delivery time
Two different questions get answered with the same word — fast — and they are not the same question. Polling cadence is how often the bot looks. Delivery time is how long it takes for something posted on X to appear in your chat. One is a setting inside our own system; the other is a sum of parts, and most of those parts belong to somebody else.
The journey has three stretches. First, X has to make a brand-new post servable to anybody at all — until it does, no tracker on earth can see it, however often it asks. Second, somebody has to notice, and that stretch is the one the cadence bounds: check more often, and the window in which a post sits unseen gets narrower. Third, Discord or Telegram have to accept the message for your destination, and a rate-limited webhook or a busy group moves that on a schedule of its own.
Only the middle stretch is ours. That is why this page states a cadence and no arrival figure: a delivery number would be a blend of our work and two other companies' work, and it would drift for reasons we could not honestly explain to you. So when you compare bots, read a speed claim carefully and ask which stretch it covers — most of them quietly measure the middle one. The longer version of that argument is on how fast crypto Twitter alerts really are.
What does Turbo change?
Turbo changes how often we look, and nothing else. Every tracked account sits somewhere in a polling rotation. On a standard plan that is the normal rotation — quick, but not the quickest schedule we run. Turbo moves the accounts on your subscription into the fast rotation, the one where we re-check a tracked account as often as every 300 ms.
What Turbo does not change is the alert itself. Same coverage — posts, replies, quote tweets with the parent shown, contract addresses on Solana, EVM and TON with DexScreener data and buy buttons, auto-translation, X-lists, profile changes, follows and unfollows, Spaces, keyword search and Bio Watch. Same destinations in Discord and Telegram. Nothing is withheld from a standard plan and unlocked by paying more; the only thing Turbo buys is frequency.
And Turbo is an add-on, not a fifth tier. There is no Turbo plan to subscribe to: it costs $9, $19, $29 or $39 depending on which plan it rides on, it is bought on top of an existing subscription, and it expires exactly when that subscription expires — so you cannot end up paying for speed on a plan that has already lapsed. The full table is on the pricing page.
What the 281 ms record is — and what it is not
Detection means the gap between the timestamp a post carries on X and the moment our system has that post in hand. Over 72 hours on production, ending 29 August 2026, we recorded 51,655 detections. The quickest of them was 281 ms.
That is our fastest measured detection. It is a single best-observed result out of fifty-one thousand, and we would be misleading you if we let it read as anything else:
- It is not typical. Plenty of those 51,655 detections were slower, some considerably. A record tells you what the system is capable of on a good pass, not what the next post will do.
- It is not a guarantee. Nothing on this page is. An account that X is slow to serve, a rate-limited response, or a burst of activity across many tracked handles at once all push a given detection above the record.
- It is not the whole journey. Detection stops the clock at the point we have the post. It says nothing about what happens after — see the next section for why we stop there.
We publish it because it is real and because it is checkable against the thing it describes. We label it a record because that is what it is.
What we deliberately do not claim
We do not publish a delivery time — a figure for how long after a post the alert appears in your chat. Not a median, not a typical case, not a range. That is a deliberate omission, and here is the reasoning.
- Half the journey is not ours. X decides when a brand-new post becomes servable to anybody at all, including us. Then Discord or Telegram decide when they accept the message for your destination. We control neither end.
- It varies, and the variance is not noise. A busy webhook gets rate-limited by Discord. A large Telegram group behaves differently from a private chat. A quiet Tuesday and a launch-day scramble are not the same system. Averaging across all of that produces a number that is true of nobody.
- A single figure would invite exactly the wrong comparison. Most published speed numbers in this category are detection on the vendor's own server. Placing a delivery figure next to a detection figure and calling it a ranking compares nothing at all, and we would rather not supply the ammunition.
Nor do we claim to beat anyone. Other tools publish figures about themselves; we have no way to reproduce those numbers under the same conditions, we do not independently verify them, and so we will not stand ours beside theirs and declare a winner. The only comparison worth anything is the one you run yourself, and it is described at the bottom of this page.
We also do not claim the floor is zero. X needs a moment before a brand-new post is visible to anybody — including us. No tool goes below that floor, whatever it advertises.
Speed only counts if the alert is complete
A fast alert that arrives without the thing you needed is a slow alert. Plenty of trackers only catch the main tweet and miss the reply where the contract address actually drops, or hand you a bare link with no chain, no price and no liquidity. X-Relay catches replies and quote tweets as first-class events and attaches contract-address data before the message is sent. A missed reply is infinitely late, and no cadence rescues it.
Check it yourself
You do not have to take a number on a marketing page on faith, and the two on this one are not the sort you can verify by staring at them. What you can verify is the thing you actually buy. The live uptime page lists every subsystem in the path and the current state of each. The 3-day trial needs no card, and Discord has a permanently free 2-account tier: point a tracker at an account that posts often, leave it for a day, and compare what landed in your chat against that account's own timeline — both when it arrived and, more importantly, whether everything arrived at all.
FAQ
How often does a Twitter alert bot check for new tweets?
It depends entirely on the bot, because the checking interval is a design choice rather than a property of X. X-Relay re-checks a tracked account as often as every 300 ms. That is the polling cadence — the gap between one look at a tracked account's timeline and the next — and it bounds how long a new tweet can sit there unseen. Accounts on a standard plan sit in the normal rotation; Turbo, a paid add-on rather than a plan, moves them into the fast one. Separately, our fastest measured detection was 281 ms, the quickest single result out of 51,655 detections recorded on production over 72 hours ending 29 August 2026 — a record, not a typical figure.
Is polling cadence the same as how fast an alert arrives?
No, and they get confused constantly. Cadence is how often we look at a tracked account. Arrival is a sum of three stretches: X has to make a brand-new post servable to anybody at all, we have to notice it, and then Discord or Telegram have to accept the message for your destination. Only the middle stretch is ours, and it is the one the cadence bounds. That is why we publish how often we check and do not publish a delivery figure.
What does 281 ms measure?
Detection: the gap between the timestamp a post carries on X and the moment our system has that post. 281 ms is our fastest measured detection, the best single result out of 51,655 detections over 72 hours on production ending 29 August 2026. It is a best case. Plenty of detections are slower, and we do not present it as what you should expect.
What does Turbo change?
How often we check a tracked account, not what we send. An account on a standard plan sits in the normal polling rotation; with Turbo it joins the fast rotation, where we re-check a tracked account as often as every 300 ms. The alert itself is identical — same coverage of posts, replies, quotes, profile changes, Spaces and keyword search, same contract-address enrichment, same translation, same destinations. Turbo is a paid speed add-on at $9, $19, $29 or $39 depending on your plan, not a fifth tier, and it runs for exactly as long as the plan it rides on.
Why do you not publish a delivery time?
Because we could not stand behind it. Once we have a post, the rest of the journey runs through systems that are not ours: X decides when a new post is servable to anybody, and Discord or Telegram decide when they accept the message. Both vary, and a rate-limited channel moves the figure more than anything on our side. So we publish what we control — how often we look — and one honest record for detection.