5 AI engines queried and scored — a missing engine (dashed) flags where you're invisible — Architect drafts campaign assets from the result in under 2 minutes.
AI Search is the New Frontier – Understand How AI Sees Your Brand
More people are asking AI for recommendations instead of googling around, and Umbra helps you understand how your brand shows up in those answers, fix what's wrong, and launch campaigns fast. Umbra is currently in closed beta with limited, invite-only spots.
A web-based dashboard — no install, just log in.
Umbra crawls your brand, scores how visible you are to AI, publishes your verified facts where the engines read them, and gives you the tools to fix gaps and build campaigns — grouped into five workspaces:
A view from the Command Center — not this site's own traffic. Each building is one destination in your account; height tracks total visits, lit windows flag recent visitors.
Real-time site telemetry on one screen — traffic rendered as an isometric city, brand health, live campaigns, and a card layout you arrange yourself. Connect Google Analytics for live sessions and conversions alongside it.
We crawl your site and score visibility across five major AI engines — GPT, Perplexity, Gemini, Claude, and Copilot — then report the technical picture too: SEO and AEO scores, sitemap, robots, and whether you're serving an llms.txt.
Surface the exposure a scan turns up on your public surface, before someone else finds it.
Everything the scans flagged, ordered by what actually moves your score — so you work a queue instead of a report.
Page-by-page detail: what each URL says about you, what markup it carries, and where the gaps are.
Track who else comes up when the engines answer questions in your category, and how their pricing and positioning shift.
Where your brand is being named across the sources these engines read.
Your identity record — name, description, industry, verified domain, social handles. A completeness meter shows what's still missing, because a thin profile makes for thin answers.
The facts the crawl found — products, FAQs, content topics — get published back to your own site as schema.org markup and an llms.txt file. That's the part answer engines actually read. You review the change as a diff and publish it yourself; nothing overwrites a hand-edit silently.
A hosted endpoint serving that same verified set to AI tools someone connects on purpose — Claude Desktop, Cursor, and anything else speaking MCP. Nine tools cover your profile, products, FAQs, content topics, schema types, assets and scan summary, plus any custom tool you write. Worth being straight about: this is opt-in and client-side. It doesn't change what ChatGPT or Perplexity tell a stranger — the published artifacts above are what do that.
Give us a short brief and get a full campaign back — objectives, messaging, audience targeting, channel strategy, ad copy — ready to launch on Meta, Google, TikTok, or YouTube in under 2 minutes. It all lands on one timeline with live and scheduled status and running budget totals, and exports when you need it elsewhere.
Generate high-fidelity product imagery, cover designs, and commercials. Upload your own product references or start from a generated one, then set the environment and lighting.
Headless server-side research that returns cited briefs and reports. Ten workflows ship with it — deep research, SEO audit, an SEO/MCP audit that reviews your own ground-truth configuration for citation gaps, competitive and market intel, lead research and gen, knowledge base, design clone, and a live co-pilot clipper. Route each run to Claude, Gemini, or a free model.
Keep the working notes and outputs from your research runs where you can come back to them.
Every creative — images, video, ad units — in one searchable library.
Under the hood: Umbra runs on Firebase Genkit endpoints inside Vercel serverless functions, paired with a custom AI model we trained specifically on conversational search and how AI engines structure their answers.
What's coming: Umbra already works great on its own, but we're building a pairing with /light, our web browser (currently in closed alpha), to add live web capture, a local data vault, and an MCP Explorer for testing how AI tools query your brand data.