Trust · AI
AI Disclosure
Last updated
Mentionwell is built on third-party large language models. This page is the plain-English version of how AI is used in the Service, what data leaves us, and what you should tell your readers.
1. What's AI-generated
The Service uses AI for the following stages of the article pipeline:
- Onboarding — turning your scanned site into a brand profile, audience description, and content taxonomy.
- Headlines — generating SEO-tuned headline candidates.
- Research — summarising sources we retrieve through Firecrawl and Exa.
- Outline + Draft — writing the article body in your brand voice.
- Editorial critic — automated pass for accuracy, redundancy, and tone.
- Metadata + FAQ — title, meta description, and FAQ Q&A pairs.
- Embeddings — vectorising articles for semantic retrieval.
- Images — generating hero and inline images for each article.
2. Providers we use
The current model providers are listed at /subprocessors:
- Anthropic (Claude) — primary text model.
- OpenAI (GPT) — secondary text model + embeddings.
- fal.ai — image generation.
3. What data leaves Mentionwell
When you onboard a domain we send to providers:
- Public content scraped from your domain (text, headings, schema, anchors).
- Brand profile, audience, and taxonomy as we synthesize them.
- Headlines, drafts, and prior articles needed for context.
We do not send: account credentials, billing info, or content from sites you do not own.
4. Training opt-out
Per provider terms in effect at the time of writing, content sent through the commercial APIs of Anthropic and OpenAI is not used to train their models. Provider terms govern in case of conflict; check the links on the Subprocessors page for the current language. fal.ai's terms similarly distinguish API usage from training data.
If a future provider has different defaults, we will list them on this page and on /subprocessors with at least 14 days' notice for B2B customers under a DPA.
5. Labelling and disclosure
You are responsible for any disclosure required by your jurisdiction or platform:
- EU: under the AI Act, providers and deployers must disclose AI-generated content where it is presented as authentic to a real person or event. We supply tooling to help with that, but the disclosure obligation rests with you.
- USA: state laws (e.g., California's AB 2655) may require disclosure for political content.
- Platforms: some advertising platforms and app stores have their own AI-disclosure rules. Check before you publish.
The Service does not auto-add an "AI-generated" badge to articles by default; you can add one via your CMS template.
6. Accuracy and limits
AI output is generated probabilistically. It can include hallucinated facts, outdated information, or misattributed quotes. The editorial-critic pass reduces this but does not eliminate it. Always review every article before publishing it to your audience. We are not responsible for content you choose to publish.
7. Your rights to the output
You retain ownership of the articles, FAQs, metadata, and images Mentionwell generates for you, subject to the terms of our Terms of Service and the underlying providers' terms. Generated text and images may be similar to output another user receives in response to a similar prompt; we make no exclusivity claim.
Questions or feedback: privacy@mentionwell.com.