Features

The drafting pipeline

Every article passes through 11 stages before it's published. Each stage has its own model strategy and prompt; each stage emits structured progress so you can watch a draft come together.

1. Research — Serper + Exa SERP + Firecrawl crawl of competitor pages, with a transparent source list.
2. Synthesize — extract key insights, AEO signals, top questions, and the unique angle.
3. Outline — search-intent-aware section plan; H2s become canonical TOC entries.
4. Draft — section-by-section writer with grounded citations.
5. Reflection — editorial critic graded on accuracy, AEO readiness, factual grounding, and brand fit.
6. Rewrite — automatic rewrite when the reflection critic scores below 0.8 confidence.
7. Metadata — meta title, meta description, excerpt, tags, category — written for CTR and search-intent match.
8. FAQ — frequently-asked-questions block + JSON-LD FAQPage schema.
9. Embed — vector embedding for dedupe + internal-link discovery on the next draft.
10. Images — featured + infographic via FAL Gemini-3 Pro 4K with a per-site visual identity prompt.
11. Render — final HTML + Markdown + JSON-LD ready for headless consumption.

AEO / GEO / LLMO + classic SEO — on every article

The pipeline isn't optional. Every article ships with:

• AEO (Answer Engine Optimization) — short, answerable subheadings; a TL;DR block; FAQ accordion; clean citation list.
• GEO (Generative Engine Optimization) — structured semantic chunks; clear topical authority signals; controlled keyword entity coverage.
• LLMO (LLM Optimization) — llms.txt + llms-full.txt on the parent site; markdown-source endpoints (/post.md) for AI crawlers; ChatGPT / Claude / Perplexity-friendly summary first.
• Classic SEO — JSON-LD BlogPosting + BreadcrumbList, sitemap.xml, RSS / JSON Feed, canonical URLs, OG tags.

Headless delivery

You choose how articles reach your site:

• Reader API — your site reads articles directly from a public bearer-token-scoped endpoint (or via the mentionwell-reader npm package against Supabase).
• Webhook push — Mentionwell POSTs every published article to your endpoint with HMAC signing.
• GitHub MDX — Mentionwell commits MDX files to your repo for static-build pipelines.
• Direct API — drop articles straight into WordPress, Webflow, Ghost, Shopify, or Notion via existing connectors.

Per-site brand profiles

Every site you onboard gets its own brand profile: tone, audience, pitch rules, competitor blocklist, image style brief, and a hand-tuned taxonomy of content clusters. The pipeline reads from this profile on every draft so a cottage real-estate site never gets a tech-startup hero image.

Per-site image style briefs

Image generation respects a per-site imageStylePrompt. Tactile claymation for a Hebrew-translation app, vivid editorial nature photography for a dive guide, dark cinematic founder portraits for an AI agency — each site stays visually consistent across hundreds of articles. All images mirror to Cloudflare R2 so you never depend on vendor CDN uptime.

Bulk operations

Imported articles from another platform? Run the bulk optimize pass to refresh metadata, inject FAQs where missing, and 2026-aware year-refresh stale references. Slugs are never changed — search rankings are preserved.

Operational dashboard

• Live drafting progress (research → render) with per-stage timing + cost
• Pipeline / Drafts / Published columns with kanban-style approval flow
• Autopilot scheduling — pick days + UTC hours; the queue handles the rest
• Cost tracking per article, per stage, per model
• Job history + per-job log streaming via Server-Sent Events

Multi-tenant for agencies

Run hundreds of sites from one dashboard. Per-site reader keys. Per-site quotas. Per-site image styles. White-label reader package. See pricing for the Agency tier.

Onboard your first site