YouTube inauthentic content policy and bulk AI video runs

YouTube's inauthentic content policy, the renamed repetitious content policy effective July 15, 2025, says content should be original and authentic, not "mass-produced, generic, repetitive, or manipulative". Its examples include AI-generated content made with generic or unoriginal templates that gives the impression of mass production. Running a batch is not the problem the page names; the sameness of what comes out is.

Quotes are from YouTube Help, read 2026-09-30. YouTube decides how it applies the policy; this post makes no prediction about any channel.

What does YouTube say, in its own words?

The page says this type of content "has always been ineligible for monetization under our existing policies, where creators are rewarded for original and authentic content". It also describes videos where each one does not deliver creative, educational or other value to the viewer. The separate reused content policy, covering commentary, clips, compilations and reactions, is unchanged.

How big can a Sume bulk queue be?

A bulk request queues "up to 100 Format runs with a concurrency window", a server-side queue of ordinary Format runs. Each item is one sandbox, one agent turn and one run receipt, and concurrency is an integer from 1 to 16. See Bulk runs.

How do I keep each item distinct?

Each entry in items is the same body as a single run, with its own instruction and input. Give every item a different subject, script and source material instead of one prompt with a number changed. The queue does not do this for you: if all 100 items carry the same instruction, you have asked for 100 near-copies, which is the pattern YouTube describes.

curl -X POST "https://api.sume.com/v1/formats/chase/product-promo/bulk-runs" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "concurrency": 3,
    "items": [
      { "instruction": "Demo of the hinge, 20 s", "input": { "url": "https://example.com/hinge.jpg" } },
      { "instruction": "Unboxing notes, 30 s", "input": { "url": "https://example.com/box.jpg" } }
    ]
  }'

What should I check before I publish?

Watch a few outputs as a viewer would and ask whether each adds something the others do not. Poll the queue and read each child run; completed means every item is terminal, not that all succeeded, so branch on counts.failed. For label rules on voiceover, see does YouTube require a label for AI voiceover Shorts; for the queue itself, Format bulk runs.