Automation Policy
Effective July 14, 2026
Automation must be declared before it acts, approved before it participates, visibly labeled, useful to the relevant Topic or Circle, and accountable to a reachable human operator.
What counts as automated
An account is automated when software selects, submits, schedules, reacts to, or materially controls platform activity without a person deliberately choosing each action. Writing assistance alone does not automatically make an account a bot, but users remain responsible for everything submitted.
Required approval
- Choose Automated during registration and explain the purpose and operator.
- Do not participate while the account status is pending or rejected.
- Keep the visible Automated label and operator contact information accurate.
- Use only approved access methods and obey route, daily, and operator-specific limits.
Allowed purposes
Approval may be considered for narrow public-interest utilities such as transparent alerts, structured public-data updates, accessibility helpers, or steward tools. Approval is discretionary and may be scoped to specific Topics, Circles, actions, or schedules.
Prohibited automation
- Synthetic community members intended to appear human.
- Mass posting, commenting, following, joining, reporting, messaging, or reactions.
- Engagement manipulation, reciprocal signaling, astroturfing, or coordinated account networks.
- Unapproved scraping, model training collection, vulnerability scanning, credential testing, or rate-limit evasion.
- Generating high-stakes claims without reliable sourcing and human accountability.
- Re-registering or rotating infrastructure to evade rejection or enforcement.
Enforcement and review
Automated activity may be delayed, rate-limited, quarantined, removed, or suspended. Duplicate fingerprints, action budgets, and risk signals support review, but AI-text detectors are not treated as proof. Operators can contact support to request approval or review.