AI Disclosure

What we use AI for

Mora Discover uses AI to do three things: cluster duplicate coverage of the same story across sources; synthesize a single read per cluster; and rewrite headlines for clarity when a source headline is misleading or paywall-locked. The model behind synthesis is Gemini 3 Flash via the Vercel AI Gateway, run with a custom prompt that requires source attribution on every paragraph and refuses to synthesize on thin or contradictory clusters.

What humans do

JT, the founder of Mora and editor-in-chief of Discover, oversees the editorial layer. That means: defining and updating the synthesis prompt; setting and enforcing the source-quality bar; running a daily review pass on a random sample of published output; filing corrections when synthesis misattributes a claim or misreads a primary source; and re-tuning the prompt every couple of weeks based on what the review pass surfaces. JT's byline appears on every Discover article because JT signs off on the editorial pipeline that produced it.

Hand-written content includes the topic-page introductions on /discover (one per topic, written by JT), this disclosure, the editorial policy, and the methodology page.

What AI does not do

AI does not pick which sources to ingest — that source list is curated and editable. AI does not write the topic-page introductions. AI does not auto-publish corrections. When the synthesis prompt fails its own attribution check or flags low confidence, the article is held back, not published with a hedge.

Why we disclose this

Two reasons. First, ecommerce operators using Discover as part of their morning content workflow deserve to know what came from a model and what came from a human — the trust calculus is different. Second, AI-generated content without human oversight is a Quality Rater Guidelines red flag (Google's 2024 update made this explicit), and we want the record clean: this feed is AI-assisted, human-supervised, with the supervision specified.

Models and infrastructure

• Synthesis: Gemini 3 Flash (gemini-3-flash-preview) via Vercel AI Gateway
• Headline rewrites: Gemini 3 Flash Lite (gemini-3.1-flash-lite-preview)
• Embeddings (clustering): Gemini text-embedding-004
• Hosting: Vercel · Supabase Postgres · Vercel Edge cache with tag-driven invalidation

Contact

Questions about AI use in Mora's pipelines, requests for additional disclosure, or suggestions on the disclosure language above: hello@mora-marketer.com with subject line “AI disclosure”.