Rules
Original Content Rewards
Xaudit does not restate the program. X’s Help Center is the source. This page covers how the tool scores and caches — not eligibility, payout, or program terms.
For You ranking and Original Content Rewards are different systems. A handle in the box runs an account audit. A draft or status URL scores one post against the published program rules.
Ranking is not payout
For You ranking and Original Content Rewards are different systems. The payout columns judge the written program only. The ranking line is a For You estimate and does not change payout quality.
Weights multiply this viewer’s P(action), not Analytics counts. Mutual +15 reply weight is originals from mutual follows only — never replies or reposts, and not applied here.
Live order: Phoenix inference → RankingScorer → VM → Top-K → visibility ALLOW / INTERSTITIAL / DROP → DedupConversationFilter.
What Xaudit scores on a post
- OriginalityHigh / Moderate / Low / Non-original
- HookHigh / Medium / Low
- DwellHigh / Medium / Low
- Format & structureHigh / Medium / Low
- Genre / nicheHigh / Medium / Low
- Meaningful addHigh / Medium / Low / n/a
- OCR-eligible typePass / Caution / Fail · written program
- Solicitation / CTAPass / Caution / Fail · written program
- Home card (≥50% visible)Pass / Caution / Fail · written program
- Home-impression fitnessPass / Caution / Fail · written program
- Program riskPass / Caution / Fail · written program
- For You ranking estimate (not paid)High / Medium / Low · ranking only
- AI-writtenHuman / Mixed / Obvious AI
- Overall payout qualityHigh / Medium / Low · written program
Terms used in a report
- Dwell. A ranking estimate of reason to stay — not a payout metric. High: original media, an article body, or enough of a take to read twice. Medium: some hold, not a linger. Low: nothing to stay for. “Not yet dwell” means a little hold, not enough that someone would linger.
- Meaningful add. Used mainly on quotes. The quoted post is not scored — only what you wrote. High: an argument, a miss, a fact, or an implication. Low: agree/disagree, “this”, a restatement, or a shrug. If the quote would still make sense without your sentence, the add is thin.
- Asks. CTAs that beg for engagement — “like if…”, “follow me for more”, “reply and I’ll follow”. Saying a post did not reach people who do not follow you is graph language, not an ask. Recommending that people follow someone else is not an ask. Repeated self-asks are a removal risk.
- Reply-loops. The same thin reply over and over (GM, thanks, emoji on every commenter). Daily identical replies are a program fail.
- Do not add asks or reply-loops under it. The post already looks like a clean program fit. Do not bolt solicitation or a reply-loop onto it. Those are program fails, not ranking tactics.
How live account audits are cached
Live handle audits are cached by the account you audit so repeated requests do not burn fresh lookups.
- Allowlisted accounts: a successful report is kept for about 39 minutes. Hitting the same handle again in that window returns the same report you already saw — it is not re-run as a public/unauthorized lookup.
- Other accounts: successful reports are kept longer (about 12 hours) so public traffic does not keep re-pulling the same handle.
- The cache key is the audited handle, not the person clicking the button. Anyone who audits that same handle inside the window gets the stored report.
- Empty or failed pulls are not treated as a successful cached report. After the window expires, the next request is live again.
- Draft / paste audits are not subject to this handle cache. Burst and hourly caps can still apply if the site is under heavy load.