01
Classify & Exclude Noise
Automatically classifies discovered URLs into page roles (homepage, category hub, product family, PDP, guide, trust page) and excludes carts, logins, search pages, and pagination noise from scoring.
Flagship Module • 8-Category GEO Report & Catalog Governance
Make Every Product Family, Spec Table & Answer Verifiable to AI Agents
AEO & Agentic Storefronts is the technical readiness core of Forge Commerce (.si). Score your storefront across our 8-category GEO Report, group variants into parent Product Families with inherited FAQs, liberate PDF specs into visible HTML & 1:1 JSON-LD, ground AI authoring in your Account Knowledge Base, and operate everything via the @tvs/forge-mcp server.
How the GEO Report Engine Works
A 4-Stage Diagnostic Pipeline That Separates Readiness Score From Evidence Confidence
Legacy SEO crawlers let 500 identical SKU pages drown out a broken homepage and conflate missing data with a low score. Forge's GEO Report fixes both architectural flaws.
02
Cluster Templates (1 + ln(N))
Groups pages sharing a structural template and samples canonicals, outliers, and high-traffic URLs. Each cluster is weighted by 1 + ln(members) so 400 PDPs never mask category or homepage gaps.
03
Deterministic + SI Judgment
Combines 60% deterministic rule checks (HTTP status, bot access, JSON-LD validity, HTML spec tables) with 40% structured LLM judgment grounded in verbatim page quotes.
04
Confidence & Priority Ranking
Computes Confidence (100 × E × A × J × T) on the server and ranks Top Improvements by impact × ease × confidence with exact page & template scope.
8-Category GEO Report Rubric
Scored 0–100 (A–F Grade) with Independent Server-Computed Confidence
Forge replaces noisy, unweighted SEO audits with an 8-category agentic readiness model using template clustering (1 + ln(members)) so 500 identical SKU pages never drown out your homepage or category hubs.
01 • Access15% Weight
Can search engines and AI bots reach this page?
HTTP 200 status, canonical coherence, indexability, and explicit crawler directives for GPTBot, ClaudeBot, PerplexityBot, and Google-Extended.
02 • Answers16% Weight
Does the page directly answer likely buyer questions?
Answer-first opening blocks, heading-to-answer alignment, visible FAQ pairs, and Product Family inherited FAQs.
03 • Facts16% Weight
Are specs, numbers, and claims concrete and quotable?
HTML spec tables liberated from PDFs (_forge_aeo_specs), units, tolerances, SKUs, and schema additionalProperty parity.
04 • Commerce12% Weight
Are offer, price, availability, and product details machine-readable?
Complete Product/Offer JSON-LD, SKU identifiers, stock/lead-time clarity, or explicit RFQ quote routing.
05 • Structure12% Weight
Is the page cleanly organized for machines and humans?
Semantic heading hierarchy, tabular data markup, clean DOM extraction, and zero parity drift between visible copy and JSON-LD.
06 • Trust11% Weight
Is it clear who stands behind this content?
Organization identity, certifications (ISO/ANSI/ESD/GMP), author/entity provenance, and verified compliance boundaries.
07 • Discovery8% Weight
Is the page connected so crawlers and agents can find it?
Internal link graph, XML sitemap coverage, llms.txt agent routing hygiene, and parent-to-variant family linking.
08 • Actionability10% Weight
Can a buyer or AI agent take the next step cleanly?
Clear conversion paths, structured RFQ/cart endpoints, and governed Forge Foundry MCP tool availability.
Flexible Execution Economics
Four GEO Report Run Modes—Including Zero-Ingot MCP Harness
Run diagnostics on Forge's managed multi-provider router, use your own API keys, or let your connected MCP coding agent perform structured page assessments at zero Ingot cost.
Platform Mode (Ingots)
One-click execution in the Forge UI using your account's Ingot balance (100 Ingots = $1.00). Shows an upfront reservation preview and automatically refunds unused Ingots on completion.
BYOK Mode (0 Ingot LLM Cost)
Connect your organization's OpenAI, Anthropic, Google Gemini, or OpenRouter API keys in Settings › AI & Usage. Forge orchestrates the crawl and rubric while billing LLM tokens directly to your provider key.
Rules-Only Deterministic
Runs 100% deterministic checks across HTTP status, bot access, canonical tags, visible HTML spec tables, and JSON-LD schema parity in seconds—with zero LLM calls required.
MCP Harness Mode (0 Ingots)
Your connected MCP client (Claude Code, Cursor, or Codex) fetches page packets via forge_geo_report_get_Input_packet, evaluates pages against Forge's JSON Schema, and submits validated results at 0 Ingots.
Catalog Architecture & Knowledge Governance
Product Families, Liberated Specs, Account Knowledge Base & Staging → Prod Promote
Built from real-world industrial and multi-brand deployments where hundreds of dimensional variants share core engineering truths.
Parent Product Families & Shared FAQ Inheritance
Stop copy-pasting the same compliance or material FAQs across 40 dimensional SKUs. In Forge, a parent Product Family owns the shared product-line FAQs and canonical thumbnail. Child SKU variants automatically inherit family FAQs in both visible HTML and FAQPage JSON-LD while maintaining their own SKU attributes.
- Single-point FAQ authoring for entire dimensional product lines
- Automatic child SKU inheritance with override & addendum support
- Zero duplicate-content drift across WooCommerce or Shopify variants
PDF-to-HTML Spec Liberation (_forge_aeo_specs)
Extract tolerances, dimensions, materials, and certifications from locked PDF datasheets or unstructured descriptions into clean HTML specification tables stored in _forge_aeo_specs and mirrored 1:1 into Product.additionalProperty JSON-LD.
- Machine-quotable HTML tables for procurement engineers & AI agents
- Strict 1:1 parity between visible page specs and Schema.org JSON-LD
- Directly improves GEO Report Facts & Structure grades
Account Knowledge Base (Settings › Knowledge)
Every AI generation in Forge—whether authoring Product Family FAQs, liberating specs, or enriching Valkyrie Assist outbound columns—is grounded in your Account Knowledge Base: Company Overview, Value Proposition, Brand Voice, Compliance Guardrails, and up to 10 supporting documents.
- Enforces brand voice and regulatory guardrails (DSHEA, ANSI/ESD, ISO)
- Shared context spine across AEO & Valkyrie Assist workflows
- Editable via UI or Operator MCP (
forge_knowledge_profile_update)
Multi-Storefront Staging → Production Promote
Connect up to 3 storefronts per account. Author and validate Product Families, shared FAQs, and liberated spec tables on your staging storefront, then run Promote to copy approved items to your production storefront as safe drafts.
- Matches families by slug and child products by SKU or slug
- Always stages as draft on production—never overwrites live copy unreviewed
- Full audit trail across UI and
forge_aeo_promoteMCP tool
Native Storefront Write Paths
Deep WooCommerce / WordPress Native Plugin + First-Class Shopify GraphQL
Compare how Forge synchronizes catalog families, liberated spec tables, and review-gated schema across your storefront stack.
Native Integration • ForgeAEO WordPress Plugin (v1.7+)
WooCommerce & WordPress: Product Families, _forge_aeo_specs & Staging → Prod Promote
Industrial manufacturers and B2B distributors frequently run deeply customized WooCommerce catalogs where engineering specifications are trapped inside downloadable PDFs and product lines span dozens of dimensional SKUs. Forge Commerce pairs with our native ForgeAEO WordPress plugin to synchronize catalog structure, parent Product Families, liberated HTML spec tables, and 1:1 Schema.org JSON-LD.
- Parent Product Families own shared FAQs and thumbnails; child SKU variants inherit family FAQs automatically alongside SKU-specific specs.
- Liberates locked PDF datasheets into visible HTML specification tables (_forge_aeo_specs) paired 1:1 with Product.additionalProperty JSON-LD.
- Multi-storefront Staging → Production Promote workflow copies approved family & product drafts without overwriting live pages unreviewed.
- Manages /llms.txt agent routing hygiene and verifies crawler access for GPTBot, ClaudeBot, PerplexityBot, and Google-Extended.
{
"productFamily": "esd-conductive-tote-bins",
"sku": "NVX-4012-ESD",
"status": "Valid",
"inheritedFamilyFaqs": 4,
"_forge_aeo_specs": [
{ "name": "Surface Resistivity", "value": "10^3 - 10^5 ohms/sq (ANSI/ESD S20.20)" },
{ "name": "Inside Dimensions", "value": "12.00 x 9.00 x 2.50 in" },
{ "name": "Material", "value": "Carbon-Loaded Conductive Corrugated Fiberboard" }
],
"jsonLdParity": {
"Product": "additionalProperty (3 items) + Offer",
"FAQPage": "4 inherited family Q&As + 2 SKU Q&As"
}
}First-Class Integration • Shopify Admin GraphQL API
Shopify & Shopify Plus: Review-Gated Metafields, Variant Clarity & JSON-LD Parity
For DTC, nutrition, and multi-brand operators on Shopify, apps and liquid themes frequently fragment product attributes across incompatible metafield namespaces or inject duplicate schema blocks. Forge connects via the Shopify Admin GraphQL API to audit templates, structure Product Families, and stage review-gated improvements.
- Ingests products, variants, collections, and custom metafields via Shopify Admin GraphQL with zero storefront bloat.
- Groups multi-variant product lines into Product Families and generates answer-first summaries and compliance-checked FAQs.
- Presents side-by-side draft diffs grounded in your Account Knowledge Base before any metafield or theme-level JSON-LD is published.
- Scores templates across all 8 GEO Report categories with log-dampened template clustering (1 + ln(members)).
{
"storefront": "shopify-plus-production",
"familySlug": "proteolytic-enzyme-complex",
"reviewGate": "APPROVED_BY_OPERATOR",
"metafieldsStaged": [
{ "namespace": "forge_aeo", "key": "answer_first_summary", "type": "multi_line_text_field" },
{ "namespace": "forge_aeo", "key": "spec_table_json", "type": "json" },
{ "namespace": "forge_aeo", "key": "family_faqs", "type": "json" }
],
"schemaValidation": "Valid (1:1 HTML-to-JSON-LD parity)"
}Stateless Streamable HTTP MCP • @tvs/forge-mcp
Operate Your Storefront & GEO Reports Directly From Claude Code, Cursor, or Codex
Forge exposes a stateless Streamable HTTP MCP server supporting protocol versions 2026-07-28 and 2025-11-25 with scoped API keys (read, write, run). Try the interactive operator tool traces below:
Forge .si GEO Report Engine — Local Interactive Demo
Benchmark Your Storefront Across All 8 GEO Report Categories
Enter your WooCommerce, Shopify, or B2B domain below to simulate template clustering, 8-category readiness scores, and your highest-leverage improvements.
0–100 Score + A–F GradeSeparate Confidence MetricLocal Demo • Nothing Sent
Or test an instant sample preset (local deterministic simulation):
AEO & GEO Report FAQ
Technical Questions on Scoring, Product Families & MCP
How does the 8-category GEO Report replace the legacy v1 Audit Report?
The GEO Report replaces unweighted page-count averages with an 8-category rubric (Access, Answers, Facts, Commerce, Structure, Trust, Discovery, and Actionability) scored 0–100 with A–F letter grades. It uses log-dampened template clustering (1 + ln(members)) so 500 identical product pages don't drown out your homepage or category hubs, and computes an independent Confidence score (100 × E × A × J × T) on the server.
What are the 4 GEO Report run modes, and what is Harness Mode over MCP?
You can run a GEO Report in four ways: (1) Platform Mode using Forge Ingots, (2) BYOK using your own OpenAI, Anthropic, Gemini, or OpenRouter API keys at 0 Ingot LLM charge, (3) Rules-Only deterministic mode for instant structural validation, or (4) Harness Mode over MCP—where your connected Claude Code, Cursor, or Codex client performs structured page judgment using Forge's JSON Schema at 0 Ingots.
How do Product Families and child SKU inheritance work?
In technical catalogs, a single product line (such as a conductive stock box or supplement line) often spans dozens of child SKU variants. Forge lets you group child SKUs under a parent Product Family that owns the shared product-line FAQs and thumbnail. Every child SKU automatically inherits the family's FAQs while maintaining its own SKU-specific dimensions and liberated _forge_aeo_specs HTML table.
How does Staging → Production Promote protect our live storefront?
Each Forge account supports up to 3 storefronts (for example, a staging site and two production brands). Your team can generate, edit, and verify Product Families, FAQs, and liberated spec tables on staging, then run Promote to copy approved items to production as safe, reviewable drafts—never overwriting live copy automatically.
Does publishing /llms.txt alone make a storefront rank in AI search?
No. Forge treats /llms.txt and crawler directives (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) as essential agent-routing hygiene, not a standalone ranking shortcut. Real citation eligibility comes from visible HTML facts (_forge_aeo_specs), answer-first page structure, and 1:1 Schema.org JSON-LD parity.
Agentic Readiness & GEO Report
Benchmark Your Storefront on forgecommerce.si
Run your free 8-category GEO Report preview below or schedule a 30-minute walkthrough covering Product Families, spec liberation, Operator MCP, and Foundry OMS.