Instagram rules: API posting limits and AI content labels

Graph API publishing runs on a rolling 24-hour quota (quota_duration 86400s) read from GET /IG_USER_ID/content_publishing_limit. Meta's own documentation contradicts itself on the ceiling: the endpoint reference states quota_total is "currently 50", while the Content Publishing guide states "Instagram accounts are limited to 100 API-published posts within a 24-hour moving period". Carousels count as one post. A flat per-account number is therefore unsafe to hardcode from either page.

What to do: Call content_publishing_limit before each publish and respect the quota_total and quota_usage it returns for that specific account. Never hardcode 50 or 100 - Meta publishes both numbers and can change either.

Instagram exposes recommendation eligibility directly to professional accounts through Account Status, reached by "tapping into Settings, tapping Account, and then tapping Account Status". It shows whether an account's content "may be eligible to be recommended to non-followers in places like Explore, Reels and Feed Recommendations", and separately whether the account itself is eligible to be recommended "in places like Search or Suggested Accounts". Decisions can be contested: "If you think we've made a mistake, you'll also be able to appeal by hitting 'Request a Review' directly from your Account Status."

What to do: Before assuming a shadowban, open Account Status and read it - it answers the question directly. If it reports ineligibility, edit or remove the flagged posts and use Request a Review rather than waiting for a penalty to expire, because none is running.

Instagram states that "Posts that go against our Recommendation Guidelines are allowed on Instagram", and that "If someone's post goes against our Recommendation Guidelines, we try to avoid showing that post where we recommend content." The post stays up and continues to reach the accounts that follow you - Feed is described as "a mix of content from the accounts you've chosen to follow, recommended content from accounts we think you'll enjoy" - while being withheld from the recommendation surfaces where non-followers discover content: Explore, Reels, Feed Recommendations, Search and Suggested Accounts. There is no notification when this happens, which is why creators experience it as a secret penalty rather than a published rule.

What to do: Treat a sudden collapse in reach with the post still live, no strike and no notification as recommendation ineligibility rather than a ban. It is assessed per post against the Recommendation Guidelines, so there is no duration to wait out - the next compliant post is judged on its own.

Instagram names the specific reel attributes it demotes: "We aim to make certain reels less visible for other reasons, such as low-resolution or watermarked reels, reels that are muted or contain borders, reels that are majority text, or reels that have already been posted on Instagram." It also confirms the demotion is often invisible to the creator, noting it is "experimenting with new notifications to help creators understand when the reach of their reel may be limited due to a watermark". This is the closest Instagram equivalent to TikTok's For You feed ineligibility for unoriginal content, and it catches the same cross-posting workflow: the easiest file to re-upload is usually the one with another platform's watermark burned into it.

What to do: Export clean from the original editor rather than re-downloading your own post from another app. Avoid letterboxing or borders, keep audio present, do not build a reel that is mostly text, and never re-post a reel that has already run on Instagram.

Meta labels AI-generated content across Facebook, Instagram, and Threads. The label was introduced as "Made with AI" and renamed "AI info" on July 1, 2024; organic AI content labelling began in May 2024. Labels are driven by industry-standard AI indicators embedded in the file (C2PA and IPTC metadata) and by creator self-disclosure, not by a general-purpose detector. Meta's own generator adds "Imagined with AI" to photorealistic images it produces.

What to do: Do not assume Meta will detect and label a clone render on its own - labelling keys off embedded provenance metadata or your own disclosure. Neither Instagram nor Facebook exposes an AI-disclosure flag through the publishing API, so disclose in the caption when the video is realistic synthetic media.