X's automated-means rule and AI video
Original Content Rewards disqualifies content created or posted using automated means. A model is not automatically a bot.
Original Content Rewards lists a disqualifier that looks, at a glance, like a ban on generated video: content is ineligible if it "was created or posted using automated means."
A model is not automatically a bot. X has not published a carve-out that says "Veo is fine" either. The bite, read against the rest of the program, is firehose posting and unoriginal output — unattended volume and other people's files — not the existence of a generator in the stack.
What the program page actually groups together
The ineligible-content list in X's Original Content Rewards article sits in one block. Automated means is one bullet among others that are about provenance and spam, not about which renderer made the pixels:
- Copied or substantially reproduced from another creator
- Downloaded from X or another platform and reuploaded (unless you are the author)
- Created or posted using automated means
- Exclusively focused on monetization coaching
- Disinformation, Community Notes, policy violations
Separately, continuous eligibility forbids using automated tools, bots or software to manufacture likes, follows, views, comments or shares, and forbids repeatedly instructing people to engage.
That is two "automated" clauses. One is about how the post got made or published. One is about how the engagement got made. Bot-manufactured engagement is an account-level kill. This page is the first clause: the file and the poster.
X's examples of original content, in the same article, include "photos or videos you created," memes and graphics you designed, and commentary that adds a perspective. A generated clip you actually made — scripted, shot as a generation, edited, captioned as yours — sits closer to "videos you created" than to "downloaded and reuploaded." A scheduler dumping fifty near-identical generations a day sits closer to "posted using automated means" plus the originality test.
X has not drawn that line in a model table. The grouping is the evidence you have. Do not invent a licence. Do not invent a ban.
Firehose posting is the part you can stop
"Posted using automated means" is the half of the clause that does not require a philosophy of AI. Unattended posting, reply bots, auto-syndication from another platform's share sheet, and any tool whose job is to publish without a person looking at the post are the obvious read.
A person generating a clip, reviewing it, writing a caption and hitting post is not that fact pattern. A pipeline that generates, captions and publishes on a timer with no review is. The difference is not the model. It is whether a person is the publisher.
If you currently auto-post the same asset to X because it is going to TikTok and Reels anyway, that is a distribution convenience and an originality problem at the same time. X wants original content on X. A watermark from another app is cheap evidence the file did not start here. Cross-posting your own work is allowed in the reupload bullet if you are the author; doing it through a bot that also sprays replies is how you pick up the automated-means clause on the way.
Cut unattended publishing on any handle you intend to monetize. Keep a person in the loop. That is the whole operational fix for the "posted" half.
Unoriginal output is the part a generator will happily produce
The "created using automated means" half is where people over-read. Taken literally, any software-assisted file is automated. Taken in context, X is running an original content program whose failure modes are copies, reuploads, and posts whose value is still someone else's.
A generator used to manufacture a copy of someone else's video is laundering. A generator used to produce a new clip from a brief you wrote is production. The program already says minor edits — crop, filter, border, watermark, speed, simple overlay — are not meaningful transformation of third-party material. Prompting "make this TikTok again, vertical" is that fact pattern with extra steps.
What actually gets you paid, on X's own list: writing, reporting, video you created, illustrations you designed, commentary that adds something. The prompt library is useful when the prompt is the authored part — a specific setting, a specific gag, a specific product — and useless when it is a generic "viral hook" applied to a stolen reference.
| Workflow | How it reads against the clause |
|---|---|
| You write a brief, generate a new clip, edit, post it yourself | Original production; not the firehose |
| You generate 40 near-identical clips and auto-post them | Firehose + template; both halves of the clause |
| You prompt a model to recreate someone else's video | Copy, regardless of the tool |
| You schedule a reviewed, authored post for a timezone | Ordinary publishing |
| A bot publishes, replies and farms engagement overnight | Automated posting and manufactured engagement |
Batch output that does not read as batch output is the craft version of the same rule. Original Content Rewards is the compliance version. A feed that is visibly one template is the signature the originality standard was written to catch, whether or not a diffusion model sat in the middle.
The neighbouring rules still apply
Clearing automated-means does not clear X's manipulated media policy. That policy is deception plus potential harm. A generated clip of a real person saying something they did not say is a problem on that document, not on this one.
X also has no composer "Made with AI" toggle. Disclosure on X is still something you write. Original Content Rewards does not replace that, and it does not pay you extra for a label.
Use a generator as a camera. Do not use it as a poster, and do not use it as a way to ship someone else's file with a new render. That is the reading that matches the article X published. It is not a legal opinion. The article is the authority; read it again if you are about to industrialise a pipeline.
FAQ
Does using an AI video model make a post ineligible?
X has not published a model-by-model ban. The ineligible bullet is content created or posted using automated means, grouped with copies and reuploads. A clip you authored and posted yourself is the fact pattern the "videos you created" examples describe. A firehose of unreviewed generations is the fact pattern the bullet is for.
Can I use a scheduler?
A scheduler that publishes a post you already reviewed is ordinary publishing. A system that generates, captions and posts without you is "posted using automated means." Keep the person in the loop on any monetized handle.
What if the model is recreating a reference video I do not own?
That is a copy with extra steps. Minor transformation does not qualify as original content on X's own list. Generate a new asset from a brief you wrote, or do not post it into a program that pays on originality.
Is engagement automation the same rule?
No. Manufacturing likes, follows, views, comments or shares is a separate continuous-eligibility ban and an account-level risk. Automated-means on the content list is about the file and the act of posting. You can fail either. Do not fail both.