How to run a giveaway on X without wrecking your reach

#Quick answer

X allows giveaways and publishes guidelines for them. The problem is that the standard giveaway playbook, follow plus repost plus tag three friends plus repost daily for extra entries, collides with two of the five things X's own promotions page asks you to discourage. Run it as one entry per person, collect entries through your mentions rather than a hashtag search, and publish the draw method before you post. The reach cost of a giveaway is mostly self-inflicted, and the suspension risk mostly lands on your entrants rather than on you.

Last updated: October 2026

#What X's guidelines actually say

There is a real page. It is called Guidelines for Promotions on X, it lives in the rules-and-policies section of the help centre, and it is about 400 words long. Most of the giveaway guides ranking for this query paraphrase it without linking it, and several add requirements that are not in it.

Here is the whole thing, compressed:

Then the part everyone skips: compliance with applicable laws and regulations is your responsibility, and X suggests you consult an attorney. X does not vet your promotion. It is telling you where its liability ends.

Note what is absent. There is no requirement for official rules, no required disclaimer saying the promotion is not sponsored by X, and no registration threshold. Those come from other platforms' policies and from sweepstakes law, and conflating the three is the single most common error in the pages ranking above this one.

#The two mechanics X asks you to discourage

The giveaway industry sells entry multipliers because they inflate the numbers that make a giveaway look successful. X's guidelines target exactly two of them.

Volume reposting. "Repost daily for an extra entry" and "most reposts wins" are the same instruction to X's systems: post near-identical content repeatedly. X's own position is that this jeopardises search quality and can cause entrants to be filtered out of search. Your giveaway post is the thing those filtered accounts were reposting, and the entries you are counting are the posts that got filtered.

Multi-account entry. X's authenticity policy allows up to ten accounts "for different, non-duplicative purposes" and prohibits operating multiple accounts that interact with the same or substantially similar content, or that post substantially similar content to one another. A giveaway with per-entry odds is a direct incentive to break that. When X enforces on a multi-account violation, its documented outcome is that the person may be asked to choose one account to keep while the rest are suspended.

Neither of these is a rule against giveaways. They are rules about duplication, and a badly designed giveaway is a duplication machine you built and pointed at your own audience.

#What entrants are risking, from X's own enforcement code

This is the part no giveaway guide covers, because it needs the code rather than the help centre. X open-sourced the For You feed stack, and the repository includes an abuse enforcement service with its rule files checked in.

In enforcement_user.yaml, the actions attached to platform-manipulation labels are explicit. A label called anchor_campaign_suspend triggers a non-permanent suspension with the policy string PlatformManipulation. A second model, inauthentic_detection_v45, has both a suspend path and a label-plus-captcha path. The SpamHighRecall label is applied with ttl_msec: 2592000000, which is exactly 30 days. There is also a spam_liveness_check_required label that triggers the human-presence test, the same mechanism covered in our note on the X liveness check.

Two honest caveats. The file header says it is mirrored from GrowthBook dynamic config, last synced 26 August 2026, and the follower thresholds in it are deliberately mocked: the comment reads "Prod uses a different follower count floor; this is a mock value to reduce gaming." So the numbers are not the live numbers. The label names, the actions and the structure are.

And the structure is the interesting part. Before any enforcement rule fires, the file short-circuits on allowlists, on a follower-count floor, and on pagerank_skipped, which exits when a credibility score crosses a threshold. Established accounts are checked out of this pipeline early. The exposure sits with new and small accounts, which is precisely the population a prize draw recruits. You are not risking much by running a sloppy giveaway. The 40-follower account that made three throwaways to enter it is.

If you want the broader picture of how reach suppression gets reported and how little of it is visible to you, the X shadowban check page covers what X's own tooling does and does not show, and visibility limiting is the accurate term for it rather than a platform state called shadowban.

#Collect entries by mention, not by hashtag search

X's guidelines make this point in one sentence and nobody repeats it: a public search may not show every entry. If duplicate-heavy entries get filtered from search, a hashtag-based draw is drawing from a list that quietly dropped contestants, and you will not know which ones.

Mentions do not have that problem. They land in your notifications timeline, they arrive in order, and they survive search filtering because you are not relying on search. The practical version:

• Ask for a reply or a quote post that mentions your handle. Both put the entry in your notifications.
• Use a contest hashtag for discovery and visibility, not as the entry mechanism.
• Screenshot or export the entry list at the close time you published, before you draw.

This also kills the hashtag-hijack problem in X's fourth guideline. If the hashtag is decoration rather than the ticket, nobody has a reason to staple it to unrelated posts.

#Picking a winner now costs money

Until 2026, the answer to "how do I draw a winner" was a free third-party picker that called the API. That era is over, and this is the most load-bearing stale fact in every ranking page on this topic.

The X API is pay-per-usage with no subscriptions and no free allowance. Reads are billed per resource returned. From X's published rate card:

Run the arithmetic on a mid-sized giveaway. Listing the reposts of one post goes through GET /2/tweets/{id}/retweets, so 5,000 reposters is 5,000 post resources at $0.005, about $25. Resolving those into users to filter by follower count or account age is a user read each, $0.010, another $50. Pay-per-usage accounts are also capped at 3 million post reads per monthly billing cycle.

None of that is expensive in absolute terms. It is just no longer zero, which is why the free picker tool you used in 2023 either disappeared, added a paywall, or now asks you to paste your own API credentials and spend your own credits. Our X API pricing breakdown has the full rate card if you are costing out a tool.

For most giveaways, the free path is better anyway: number the entries from your mentions list, draw with a seeded random number, and show the seed. It is auditable in a way "we used a tool" never is.

#The legal line you cannot skip

X's guidelines hand this back to you, so handle it once and reuse it.

The core test is consideration. The FTC's business guidance states it without hedging: "If a purchase or payment of money is required for eligibility for a prize, it is not a prize promotion; it is a lottery, which is generally unlawful under federal and state lottery laws." That is the entire reason every real sweepstakes carries a no-purchase-necessary line. It is not a platform rule and it is not decoration.

So: publish rules on a page you control, with eligibility, dates, the prize, how the winner is drawn, and how they will be contacted. Link that page from the giveaway post. Say that no purchase is necessary. If the prize is large, if you are gating entry behind a paid product, or if you are operating across jurisdictions, get actual legal advice, which is what X's own page tells you to do.

#A mechanic that survives all of this

Put the constraints together and one shape falls out.

One entry per person. Entry is a reply that says something. Ask the question you actually want answered: what are you stuck on, what would you use this for, what did you ship this week. Follow if you want, as a condition you do not police. A contest hashtag for discovery. Winner drawn from the mention list on a published date, with a published method.

That mechanic has no duplicate posting, no incentive to spin up accounts, no dependency on search, and it produces the thing a giveaway is supposed to produce and almost never does: a reply thread with real content in it, which keeps earning reach after the prize is gone. Compare it with the usual outcome, a repost count and a follower spike that unwinds within a fortnight.

Which it will. A prize-driven follow is a weak follow, and why you lose followers on X covers the churn pattern: people who followed for a reason that has expired leave, and they leave fastest in the week after. Plan for the audience you keep, not the peak.

If the appeal of a giveaway is that it is a shortcut to distribution, the honest comparison is against doing the unglamorous version properly. How to get followers on X is the slower route, and X-Autopilot exists because the slow route is mostly a volume and consistency problem that a tool on your own machine can carry for you. It runs in your browser on your own connection, which is a smaller detection surface than a cloud scheduler posting from a shared IP pool, though no setup removes risk entirely.

#After it ends

Three things, in the week after the draw:

Announce the winner publicly in the original thread. It closes the loop for everyone who entered and it is the only proof your next giveaway will have.

Check whether your reach recovered. A giveaway post with thousands of low-context reposts changes what your account looks like to the ranker for a while. If your normal posts are flat for a few days afterwards, that is the cost, and it is temporary.

Do not run another one next week. The same audience, the same mechanic, the same duplicate-ish engagement pattern, repeated on a schedule, is the profile X's authenticity policy is written about. A quarter is a reasonable floor between them.

A giveaway is a legitimate and genuinely useful tool. It just has a narrow correct form, and the industry that sells giveaway tooling has spent a decade selling you the other one.