What it does

x-account-slop-check takes an X/Twitter handle and answers one question: is this account posting AI-generated content? 

{
  "username": "agentutility_ai",
  "max_posts": 10
}

returns a verdict for each recent post plus a rolled-up read on the account: 

{
  "posts_checked": 8,
  "posts_too_short": 2,
  "slop_rate": 0.25,
  "per_post": [
    {
      "id": "1234567890",
      "verdict": "likely_human",
      "probability_ai_generated": 0.08,
      "snippet": "just spent the morning debugging a flaky test, turns out it was a timezone thing again"
    }
  ],
  "account_verdict": "mostly_human"
}

Why

This isn't the same job as x-account-authenticity. That endpoint scores bot-likelihood from account metadata — profile age, follower/following ratio, homoglyph impersonation risk — and blends one AI-text signal across the whole timeline into a single score. It answers "is this account probably a bot."

x-account-slop-check answers a narrower, more literal question: reading the actual text of this account's recent posts, how much of it reads as AI-generated, post by post? Buyers who want a per-post breakdown — not one averaged number — were going to have to fetch the timeline and call the AI-content detector themselves, post by post. This collapses that into one call.

How the detector pass works

ai-content-detector takes one block of text and returns one verdict for that text. There's no batch mode that scores multiple items inside one call and hands back separate verdicts per item. Since per-post verdicts are the entire point here, x-account-slop-check calls the detector once per qualifying post (posts under 100 characters are marked too_short and excluded from the rate) rather than concatenating posts into one prompt and reporting a single blended score. Detector calls are chunked to cap concurrency at 10 in flight at once.

Degradation

user-tweets is the required leg — if the fetch for the handle fails, the call fails. The per-post detector pass is optional: any post the detector can't score comes back marked degraded in per_post instead of failing the whole response, and degraded: true flags it.

account_verdict reads mostly_human under a 0.25 slop rate, mixed between 0.25 and 0.6, mostly_ai above 0.6, and insufficient_data when fewer than 3 posts qualify.

Price: $0.05.