Control Room Field Guide

Agntbase Field Map

A practical guide to the business fields Agntbase maintains for Agentic GEO: entity, claims, sources, freshness, conflicts, permissions, safe actions and the files that expose them to AI systems.

Use it for client delivery Open Control Room

Glossary

Every field has a job.

Meaning, owner, evidence and output are explained in plain language.

Agentic

Agents need boundaries.

The profile helps agents evaluate, compare, route and hand off without guessing.

Roles

A new specialist workflow.

Useful for GEO teams, technical SEO, entity managers and AI business operators.

Report

Client work becomes visible.

Sources, risks, next actions and installable files become a clean client scope.

Short version

A maintained AI-facing business layer.

The files are interfaces. The product is the maintained profile, evidence, control room and agent-ready business layer behind them.

Resolve the entity

Name, domain, locations, services, identifiers and relationships must point to the same business.

Label the evidence

Facts are separated into verified, public, inferred, stale and owner-review states.

Set agent boundaries

Claims, permissions and safe actions define what can be said, routed or escalated.

Export the client pack

The work becomes an install bundle, report, approval queue and recurring maintenance plan.

Managed fields

The fields clients actually need.

This is the checklist for a GEO specialist, agency, technical SEO team, entity manager or AI business manager who has to turn a messy website into a reliable machine-readable profile.

Business entity

The company AI should resolve

Canonical name, domain, legal or public identity, locations, service areas, category, same-as links and entity relationships.

Owner: Client owner plus specialist review.
Evidence: Official website, company records, owner confirmation and canonical profile.
Output: company-profile.json, hub profile JSON, JSON-LD and Control Room.

Canonical facts

The source-of-truth profile

Short description, services, offers, audience, locations, contacts, hours, pricing notes, trust signals, limits and support paths.

Owner: Client confirms; specialist structures and publishes.
Evidence: First-party pages, intake answers and owner-approved updates.
Output: Canonical profile, website file package, client pack and install guide.

Claims

What AI may safely say

Approved claims, restricted claims, forbidden claims and claims that require owner approval before a specialist publishes or sends them to a client.

Owner: Specialist drafts; client approves sensitive claims.
Evidence: Claim policy, source links, screenshots and explicit approval notes.
Output: claims-policy.json, report notes and owner approval queue.

Sources

Why a claim is trusted

Official pages, machine-readable files, public listings, reviews, social profiles, mentions and third-party references with priority and caveats.

Owner: Specialist maps sources; client confirms official priority.
Evidence: URLs, timestamps, source type, freshness and support notes.
Output: source-links.json, evidence register and monthly monitoring.

Freshness

Whether facts are still current

Last checked dates, stale fields, changed pages, missing values, old prices, outdated offers and scheduled review tasks.

Owner: Control Room tracks; specialist reviews monthly.
Evidence: File timestamps, crawl checks, change notes and source status.
Output: Control Room, client repair queue and update history.

Conflicts

Where AI may get it wrong

Contradictions between official pages, directories, old content, reviews, cached profiles or machine files that can make AI infer the wrong answer.

Owner: Specialist flags; client resolves business truth.
Evidence: Side-by-side source comparison and owner decision.
Output: Conflict list, owner approval queue and repair tasks.

Files

The machine-readable surface

Installable files and endpoints that make the official profile easier for crawlers, assistants and agents to read without guessing from page copy alone.

Owner: Specialist prepares; website team installs.
Evidence: Reachability checks, syntax checks and file verification.
Output: llms.txt, agenthub.json, profile JSON, JSON-LD and bot hints.

AI Search Scenarios

Questions customers ask AI

Real buying, comparison, routing and trust questions checked against the profile so the specialist can see what facts are missing or unclear.

Owner: Specialist writes scenarios; client confirms business intent.
Evidence: Scenario templates, profile coverage and missing-field notes.
Output: Scenario coverage, report sections and monthly repair tasks.

Permissions

Boundaries for AI systems

What can be quoted, what needs approval, what cannot be changed automatically and which paths are safe for public agent reads.

Owner: Client sets business boundaries; specialist encodes them.
Evidence: Approval notes, permission file, legal or operational limits.
Output: Permissions JSON, agent manifest notes and claims policy.

Safe actions

What an agent may route

Contact, booking, quote, checkout handoff, support and sales paths with required fields, limits, fallback rules and human confirmation points.

Owner: Client approves; specialist turns it into an action map.
Evidence: Public forms, support policy, checkout rules and handoff steps.
Output: agent-actions.json, Agent Path Map and Control Room checks.

Client pack

The deliverable a client understands

A claim-safe package that turns technical work into an executive summary, evidence register, install bundle, approval queue and maintenance plan.

Owner: Specialist exports; client reviews and approves next steps.
Evidence: Profile data, file checks, source register and scenario coverage.
Output: Downloadable client pack, install guide and recurring task list.

Portfolio

How agencies scale the work

Multiple client profiles, shared review rules, recurring freshness checks, owner approval gaps and portfolio-level repair priorities.

Owner: Agency team manages; each client owns its facts.
Evidence: Profile status, update history, file verification and source conflicts.
Output: Control Room portfolio, agency reports and monthly service plan.

Honest limits

No ranking promises.

Agntbase does not promise that ChatGPT, Google AI Overviews, Perplexity or any other model will rank, cite, recommend or prefer a business. It prepares and maintains clearer official data, evidence and action boundaries for machines to read.

It does not train third-party AI models.
It does not control model behavior or citations.
It does not let public agents mutate profiles or charge customers autonomously.
It keeps verified, public, inferred and owner-review facts separate.

New role

Entity manager, AI business manager, Agentic GEO specialist.

The name may change. The work is already visible: someone has to maintain the official version of a business for AI systems, agents, search surfaces and client teams.

Entity Manager AI Business Truth Manager Agentic Readiness Consultant AI Discoverability Architect

PriorityWhat to fix first

Next actionWho owns the update

EvidenceWhich source supports the change

ExportClient pack, install guide and maintenance plan

Practical next step

Use the field map as the client scope.

Start with one client. Build the profile, label the claims, map sources, install files, define safe actions and turn the gaps into monthly work.

Request founder packs Open Control Room