Documentation
AIO.GEO product documentation
Everything a buyer, agency, or operator needs to understand the product: what we measure, what we refuse to claim, how the loop works, and how to run tools safely. Public method, math names, and open-core packages live here. Closed-core solver source stays private (Recipe vs Kitchen).
Grok-native operator stack
AIO.GEO Protocol is an independent product from Digital Currensy Inc. Engineering was built with Grok Build. Day-to-day operators run Cursor with the @aio-geo/mcp-server MCP and the @aio-geo/cli terminal. We do not claim official xAI endorsement. We claim measurable structure, dry-run fixes, and receipts.
Sticky "Created with Grok" chrome stays off on the commercial site by default. Attribution lives here, in footer trust links, and in llms.txt for agent discovery.
What AIO.GEO is
Closed-loop AI Search Readiness for brands and agencies.
AIO.GEO measures whether generative answer engines can find, read, and trust your site; produces ranked fix packs; applies them with a dry-run first; rescores; and can seal the before-and-after with a verifiable receipt.
We do not sell LLM rankings. Nobody can measure them honestly. We sell what can be measured: structure, accessibility, and receipts to prove it.
Primary buyers: agency owners and growth freelancers packaging $5k-$15k/mo AI visibility retainers, and B2B marketing leaders who need clear share-of-mind language without fake ChatGPT ranks.
Products & surfaces
Web audit
Free light audits, bot matrix, inspect pages, and competitor-style teardowns for CMOs.
Run an audit →Autopsy
Paste an AI answer. See cited domains and missing anchors without inventing ranks.
Open autopsy →CLI & MCP
Same engine in the terminal and Cursor/Claude Code. Dry-run apply is the default.
Operator docs →Agency control room
Multi-client roster, digests, and white-label exports for retainer delivery.
Agency portal →Proof seals
HMAC receipts over recorded experiments. Clients verify integrity at /proof/:id.
Proof example →Pricing
Free light audit. Pro $49/mo. Agency $199/mo. Live probes metered under a spend cap.
See pricing →Scoring method (public)
Versioned, explainable, printed on every report.
Methodology geo-heuristic-v3.5 produces a 0-100 readiness score from five pillars, then applies multiplicative penalties when the crawl fails hard (bot walls, blocked AI crawlers, HTTP failure, thin extractable text).
| Pillar | Weight | What it checks |
|---|---|---|
| Technical Accessibility | 25% | Crawl reach, AI bots, robots, TLS/HTTP health |
| Content Structure | 25% | Headings, answer-first blocks, extractable text |
| Structured Data | 20% | JSON-LD Organization/WebPage/FAQ, sameAs |
| Authority Signals | 20% | On-page anchors, outbound sources, entity clarity |
| Presentational Quality | 10% | Landmarks, lists/tables, alt text, text density |
Implementation details that do not change the public contract (internal clustering, probe scheduling, research prototypes) are intentionally not published here.
Optimal transport vector matching
Public mathematical authority for geo-heuristic-v3.5. Equations are the recipe, not the kitchen.
The scoring engine names entropy-regularized optimal transport distance between Abstract Syntax Tree (AST) DOM node distributions and structural target profiles. Enterprise buyers see deterministic computer science, not unmeasurable rank claims.
P* = argmin_{P ∈ U(a,b)} ⟨P, C⟩ + ε H(P)Readiness-style mapping (structural distance sketch): S_Readiness = 100 · (1 − ⟨P*, C⟩). Not a ChatGPT rank and not a citation share product.
H_BPE(X) = − Σ_x P(x) log₂ P(x)
BPE token entropy is a published content-structure signal name. Executable extractors and solvers remain closed-source. Full doctrine: docs/ARCHITECTURAL_GUIDANCE.md.
Open-core distribution and IP boundary
Recipe vs Kitchen: public formulas and npm operator packages; private platform core.
Blue-chip pattern: Google published PageRank and Attention math while keeping the live crawler private. AIO.GEO publishes method names and formulas while keeping solvers, DB handlers, and commercial UI closed.
Terminal auditor and dry-run apply runner.
npx @aio-geo/cli audit example.comModel Context Protocol for Cursor and Claude Code.
npx @aio-geo/mcp-serverProgrammatic TypeScript client.
GitHub Actions score gate (real scores only).
Private monorepo → Vercel. Public npm for operator packages. Declaration: docs/DECLARATION_MANIFEST.md
Closed loop
Measure → fix pack → dry-run apply → rescore → optional seal
01 Measure
Crawl and score with a versioned method.
02 Fix pack
Ranked, machine-readable patches with plain-English gaps.
03 Dry-run apply
Preview diffs / PR JSON. No live writes by default.
04 Rescore
Same method version; record the delta.
05 Seal (optional)
HMAC receipt for client verification.
06 Digest
Agencies export white-label summaries on a schedule.
Enterprise utility suite
First-party operator tools at /tools. No third-party iframes. No fake ranking claims.
- Scanner + fix packs
Home audit and dry-run remediation previews
- Citation autopsy
Paste LLM answers; extract who was cited
- CLI / MCP / SDK
Operator plane for terminal and Cursor
- llms.txt generator
Agent-facing site map Markdown
- Competitor duel
Side-by-side readiness pillars
- CMS paste guides
WordPress, Shopify, Webflow, Next.js
- Ticket exporter
Jira / Linear ready Markdown from fix packs
- JSON-LD builder
Organization and FAQ graph drafts
- GSC topical map
Cluster Search Console demand into H2 pillars
- robots.txt synth
Toggle AI crawlers allow vs block
- Copy optimizer
Chunkability 35 to 65 word guidance
- Unfurl simulator
Social and citation card previews
- PNG social card
1200×630 share graphic from scores
- AI policy draft
Citation open / training reserve template
- Webhook exporter
Slack and Discord audit payloads
- Proof seals
HMAC verify surface at /proof/:expId
Standing refusal: we do not sell LLM rankings. We sell structure, accessibility, and receipts.
Safety defaults
- Dry-run is default on every mutation path (web, CLI, MCP).
- Live GitHub / CMS writes require your tokens; keys are never logged.
- SSRF firewall on outbound crawls and webhooks.
- Optional live probes respect a hard daily spend cap.
- DEMO packs are labeled; never sold as measured engine share.
- Never-words in product UI: ChatGPT rank, guaranteed citations, dominate AI search.
CLI & MCP (operator)
First-party packages. Same engine as the web audit.
# Audit (structural readiness) npx @aio-geo/cli audit example.com # Fix pack preview (dry-run default) npx @aio-geo/cli apply example.com # Rescore after you ship changes npx @aio-geo/cli rescore example.com # Verify a proof seal npx @aio-geo/cli verify-proof <expId> <hmac>
MCP server exposes the same tools to Cursor and Claude Code. Configure from the CLI page. Marketers never need this surface; operators live here.
Agencies
Use the control room to register client domains, run baseline scores, export white-label reports, and send weekly digests. Sell retainers on measurable readiness and sealed deltas, not invented ChatGPT ranks.
Open agency portal →Glossary
Buyer-safe terms. Search filters this list.
AI Search Readiness
A 0-100 structural score of whether AI systems can find, read, extract, and trust your pages. It is not a ChatGPT rank and not a promise of citations.
geo-heuristic-v3.5
The published methodology version printed on reports. Pillar weights and penalty categories are fixed for a given version so rescores are comparable.
Pillars
Five weighted dimensions: Technical Accessibility (25%), Content Structure (25%), Structured Data (20%), Authority Signals (20%), Presentational Quality (10%).
Fix pack
A ranked set of concrete changes (for example robots.txt, JSON-LD, answer-first headers). Shown as a preview first. Live writes need your credentials.
Dry-run
Default safety mode for apply paths. You see the exact diff or PR payload with zero production writes until you explicitly dispatch.
Rescore
Re-running the same methodology version after changes to measure the score delta. Demos that invent jumps without a crawl are not valid receipts.
Causal Proof Seal
An HMAC-signed before-and-after receipt (T0 → T14 window when recorded) that a client can verify. Integrity of the experiment record, not an LLM ranking product.
Bot matrix
How major AI crawlers (for example GPTBot, ClaudeBot, PerplexityBot) are allowed or blocked in robots.txt and related signals.
Autopsy
Paste an AI answer and extract cited domains, entities, and missing positioning anchors so you can see who is already being quoted.
Agency control room
Multi-client roster, digests, and white-label exports for agencies selling AI-visibility retainers.
Operator plane
CLI and MCP tools for developers: audit, apply (dry-run), rescore, and verify-proof from the terminal or Cursor/Claude Code.
DEMO label
Any pack that is not backed by live engine keys or production data is labeled DEMO. Demo is never sold as measured share of voice.
Spend cap
Hard daily budget on optional live visibility probes. Missing keys fail closed; free structural audits still run.
Standing Refusal
We do not sell LLM rankings. Nobody can measure them honestly. We sell what can be measured, and the receipts to prove it.
Recipe vs Kitchen
Public equations and method names are the recipe. Private solvers, AST heuristics, databases, and commercial UI are the kitchen. Publishing the recipe does not open the kitchen.
Sinkhorn Optimal Transport
Entropy-regularized optimal transport used as a named structural matching primitive. Public docs show the textbook form P* = argmin ⟨P,C⟩ + ε H(P). Executable solvers stay closed-source.
Open-core
MIT operator packages on npm (@aio-geo/cli, mcp-server, sdk, action) plus a proprietary commercial platform core. Vercel deploys from the private monorepo.
FAQ
Can AIO.GEO show my ChatGPT ranking?
No. AI engines do not expose rankings. Anyone selling a ChatGPT rank is guessing. We measure AI Search Readiness instead.
Will you change my website without permission?
Never. Dry-run is default. Live writes need your credentials. Most teams ship fixes as PR-ready change sets.
What is a Causal Proof Seal?
A tamper-evident HMAC receipt over a recorded before-and-after experiment so a client can verify integrity independently.
Is DEMO data sold as live engine share?
No. DEMO packs are labeled. Live probes need keys, budget, and are fail-closed without fuel.
Next step
Run a free audit, or talk to ops about an agency pilot.