AI Content Policy
Content Policy
Guidelines for AI-generated content in ContentPulse
Last updated: April 2026
Next review: July 2026
Sections
Overview
ContentPulse generates AI-powered content variants from a single source piece. This policy defines what content we generate, our principles for generation, and what we explicitly will not create.
What ContentPulse generates:
- • Platform-native posts: LinkedIn, X / Twitter, Threads, Bluesky, Reddit, Facebook, TikTok scripts
- • Long-form pieces: Newsletter, blog post, Substack, Medium
- • Visual scripts: Carousel scripts, quote graphics, short-video scripts
- • Audio scripts: Podcast intros, episode summaries
- • Email sequences: Welcome series, nurture flows, launch sequences
- • Voice profiles: Per-client voice signatures learned from engagement
Generation Principles
Source-Grounded Variants Only
Every generated piece is derived from agency-supplied source material. We do not synthesize content from outside the source.
Voice-Faithful Adaptation
When a voice profile exists, generation adapts to that voice — never reshapes the agency's message.
No Fabricated Quotes
We do not invent quotes from named individuals. If the source contains a quote, we may surface it verbatim or paraphrase it with attribution.
No Medical / Legal / Financial Advice as Expert Opinion
We do not generate medical, legal, or financial advice presented as expert opinion. Source material that contains such advice is preserved with the agency's own framing.
Article 50 Disclosure on Every Surface
Every generated piece is labeled “Generated by ContentPulse” on the public share page, in the review queue, and in the composer output.
What We Will Not Generate
Political Disinformation
Content that fabricates political events, statements, or positions; content designed to mislead voters or distort the public record.
PII Fabrication or Doxing
Content that fabricates personal identifiers (addresses, phone numbers, emails, financial details) or attempts to expose private information about identifiable individuals.
Impersonation Without Consent
Content written in the voice of a named individual (other than the agency's own contracted client via A+ Ghostwriting) without that individual's consent.
Discriminatory or Hateful Content
Content targeting protected classes (race, gender, religion, national origin, age, disability status, sexual orientation).
Sexually Explicit or Violent Content Targeting Minors
We do not generate sexually explicit material, and we do not generate any sexual or violent content involving minors.
Human Oversight
All AI-generated pieces flow through the agency's review queue. Nothing reaches a publishing platform without explicit human approval and a manual publish action.
Generation
AI generates platform-native pieces from the agency's source input
Review Queue
Every piece appears in the agency's /queue, labeled “Generated by ContentPulse”
Edit / Approve / Reject
Agency owner can edit, regenerate, approve, or reject every piece individually
Manual Publish
Approved pieces are published manually by the agency — ContentPulse does not auto-publish to any platform
Full Agency Control
Agencies retain full control over what gets published, when, and where. ContentPulse generates the variants; the agency publishes the ones that earn approval.
Data Handling
Data We Process
- • Agency-uploaded media (audio, video, transcripts, URLs)
- • Generated content pieces and voice-profile artifacts
- • Engagement metrics from the agency's connected publishing accounts
- • Agency owner email + account metadata for communications
Data We Do NOT Store
- • Raw audio files after transcription completes (kept only during processing)
- • PII of an agency's end-clients beyond what the agency explicitly uploads
- • Biometric data, demographic data, or protected-class data
- • Health, employment, or financial-services records
Retention Policy
Agency data retained for 24 months after subscription termination; deletion available on request. GDPR right-to-erasure (Art 17) honored within 30 days via the in-app data-export + delete endpoint.
EU Hosting
All data processed and stored in Frankfurt, Germany (Supabase / AWS eu-central-1) for GDPR compliance.
Accuracy & Limitations
Generated Content is a Draft, Not a Decision
Every variant is a draft. Agency review is required. Voice scoring and content multiplication scores are heuristics, not guarantees of performance.
Transcription Accuracy Varies
Deepgram Nova-3 transcription quality depends on audio input quality and speaker diarization. Heavy accents, background noise, and technical jargon may reduce accuracy.
Voice-Profile Maturity
Per-client voice profiles need at least 5 published-and-engaged pieces before recursive rebuilds produce stable signatures. Earlier outputs may diverge from the agency's long-run voice.
Language Support
v1.0 is English-first. Other languages supported on a best-effort basis; voice-profile fidelity in non-English languages is below v1.0's English baseline.
Disputed Outputs
If an AI-generated output is inaccurate, biased, or violates this policy, agencies can flag it directly from the review queue.
Correction Workflow
Flagged outputs reviewed by support within 48 hours. Corrections applied to future generation prompts and voice profiles where applicable.
Feedback Loop
All corrections logged and used to improve generation behavior.
Escalation Path
Anyone — agency users, end-clients of an agency, or third parties — may dispute any AI-generated output and request human review.
Contact for AI-Generated Disputes
Subject: “ContentPulse AI Output Dispute” + the public share-page URL or piece ID
Response Time
48-hour acknowledgment; 5-business-day resolution target
Right to Human Review
Any party may request human review of any AI-generated piece originating from a ContentPulse agency. A member of the Aurora team will manually re-evaluate the piece against this policy.
Review Schedule
Quarterly Content Review
Generation prompts, refusal classes, and voice-profile artifacts evaluated for accuracy and bias every three months.
Annual Bias Audit
Comprehensive evaluation of fairness across client segments, languages, and platform-format combinations.
Changelog & Transparency
All policy updates, prompt changes, and bias-assessment results logged and publicly disclosed on the Model Card review cycle.
Last Updated
April 2026
Next Review
July 2026
For more details on our AI systems and oversight, see our Model Card.