AI Content Policy

Last updated: 27 May 2026

This policy explains how novaverb uses artificial intelligence to generate content, what we do (and don't) do with your data when you invoke AI features, and what's expected of you as the publisher of any AI-assisted output.

1. What AI features we offer

novaverb integrates large-language-model (LLM) APIs to power the following features:

• Content briefs — generate outlines, target-keyword maps, and competitor-gap notes for a topic
• Article drafts & rewrites — produce or rewrite paragraphs/sections based on a brief or existing text
• Citation analysis — query AI search engines (ChatGPT, Perplexity, Gemini, Claude) to track when your domain is cited
• Schema markup suggestions — propose JSON-LD blocks for entity coverage
• Title & meta-description rewrites — short generative variants

2. Providers we use

Currently supported LLM providers (operator-configurable per workspace):

• OpenAI — GPT-4o, GPT-4o-mini, GPT-3.5-turbo
• Anthropic — Claude Opus, Claude Sonnet, Claude Haiku
• Google — Gemini 2.5 Pro, Gemini 2.5 Flash

Each provider operates under its own terms. novaverb uses provider APIs with "do not train on customer data" flags where available; specifically:

• OpenAI API calls include the X-OpenAI-No-Training header (where supported)
• Anthropic API default is non-training on customer data (their published policy)
• Google Gemini API enterprise endpoints retain data for a maximum of 24 hours for abuse monitoring, then delete

3. What data we send to providers

When you invoke an AI feature, the following may be transmitted to the chosen provider:

• Your prompt or instruction text
• Content you select for processing (titles, headings, paragraphs)
• System context: project domain, target keywords, locale
• Length / style parameters you select

We do not send:

• Account passwords, payment information, or other credentials
• OAuth tokens, API keys, or session identifiers
• Data from other workspaces in your account
• Server logs or infrastructure secrets

4. AI output disclosure

novaverb-generated content carries a hidden metadata flag (generated_by: novaverb-ai) in the database so you can:

• Filter and review AI-assisted content separately
• Add a visible disclosure on your site if your local law requires (e.g., FTC endorsement guides, Italian AI labeling rules)
• Demonstrate provenance to fact-checkers, partners, or regulators

We do not automatically render a public "this content was AI-assisted" badge on your published pages — that's your editorial choice. We recommend visible labeling for journalistic, educational, or YMYL (your-money-your-life) content.

5. Your responsibilities as publisher

By using AI features you agree that:

• You will review AI outputs before publication for factual accuracy, originality, and tone fit
• You will fact-check claims, quotes, citations, and statistics — AI may hallucinate any of these
• You will not represent AI outputs as the work of a specific human without that human's consent
• You will comply with applicable AI labeling laws in your jurisdiction (notably the EU AI Act for high-risk uses, the Italian AI disclosure rules, and FTC endorsement guides where applicable)
• You retain liability for any defamatory, infringing, or unlawful content you publish, regardless of whether AI assisted

6. Content novaverb will NOT generate

Our internal safety filter blocks AI generation of:

• Sexually explicit content involving real people without consent
• Step-by-step instructions for illegal activity
• Targeted harassment, doxxing, or impersonation of identifiable individuals
• Self-harm content, beyond clinical / preventive information
• Medical, legal, or financial advice presented as authoritative (the AI may discuss topics, but with explicit "consult a professional" framing)

If you find these filters too restrictive for legitimate use cases (e.g., adult publishers, legal research, harm-reduction content), contact us — we may grant scoped exceptions on a case-by-case basis.

7. Copyright & ownership

You own AI outputs you generate via the Service, subject to:

• The provider's terms (most providers grant the user broad ownership; some retain training rights unless you opt out)
• Copyright limits in your jurisdiction — U.S. Copyright Office has stated that purely AI-generated works are not copyrightable without "human authorial contribution"
• Third-party rights — AI outputs may incidentally resemble copyrighted works in training data; you're responsible for ensuring your published use doesn't infringe

8. Cost transparency

Each AI feature call consumes credits from your subscription quota. Per-feature costs are displayed before you submit. Administrators can review aggregate AI spend in Admin → Cost & Usage.

9. Changes

This policy may change as AI capabilities + regulations evolve. Material changes will be announced via email + in-app banner for 30 days.

10. Contact

Questions about AI usage, training opt-out, or safety filter exceptions:

• Email contact@novaverb.com with subject AI Policy