agntbase / control room
Classic

Sign in to Control Room

Use Google sign-in after starting a trial, or sign in to an invited private workspace. The public demo stays read-only.

Checking Google sign-in configuration.
Workspaces/No organization selected

How AI sees this business

Canonical record, evidence chain, files, agent permissions and repair queue.
What is wrong with the AI picture of this business? Waiting for account data.
Trust & Visibility Index · what AI may use about this business right now
AI understanding cases · capability-based view, not just metrics what AI can understand, verify and safely use

Claim chain

One screen for the claim chain: canonical facts, supporting evidence, freshness, conflicts and agent boundaries.

UX layer
Claim chaincanonical -> evidence -> owner -> agent
Agent boundarysafe defaults before action

Create a separate client profile

Use this when the site is a different business. It creates a separate workspace, primary website, source scan, files check, recommendations and future public profile page.
separate workspace
1. New profileWebsite, Instagram, LinkedIn or Google Business becomes the primary source.
2. Separate scanServer reads that business independently from the current profile.
3. Own control roomScore, source graph, files, gaps and recommendations stay scoped to this profile.
4. Verify to editOwnership unlocks canonical changes and confirmations.
This does not add a source to the current profile. It creates a separate client profile.

Canonical profile

No draft loaded.

The official profile AI should trust

This is the clean business description we want AI assistants to reuse instead of guessing from scattered pages, old snippets, directories or social bios.

  • For SEO teams: it becomes the source-of-truth behind AI visibility work.
  • Next step: confirm what the business does, serves, offers and should not claim.

Tracked sources and sites

source graph

Where AI may learn facts from

Sources are websites, directories, social profiles and mentions that can shape AI answers. We track them so the profile is based on evidence, not vibes.

  • Why it matters: wrong third-party pages can override the official story.
  • Next step: scan official pages first, then add important mentions or stale sources.
Add a public resource to this profile Use this for evidence, mentions, social profiles or monitored sites that should stay attached to the selected profile. For another business, create a new client profile instead.
Source inboxcandidate facts, evidence state, review actions
Source controlcanonical hub, counter-evidence, source weight and reputation
Scan tracewhat the scanner checked, what it found and where it stopped
Extraction intelligenceadapter plan, source graph, quality score and owner questions
Tracked sourcescurrent source map used by this draft
Source reputation modeltransparent weighting, not fake certainty

Entity graph

Company, brands, people, services, locations and sources as connected facts.
relationship layer

Make the business easier to resolve

The profile says what the company is. The graph says how its parts connect, what supports those links, and where agents should not guess.

  • For complex clients: parent brands, locations, people and offers stop floating as disconnected pages.
  • Next step: add the important relationships and attach proof URLs where possible.
Graph healthentities, relationships, proof and conflicts
Entitiescanonical nodes this profile can maintain
Relationshipssource-backed edges for AI resolution

Why AI may be confused

Source conflicts and missing-field gaps translated into likely AI risks.

Find contradictions before AI repeats them

This tab separates real source disagreements from gaps like missing pricing, availability, contacts or proof.

  • For clients: this turns “AI got us wrong” into a fixable evidence task.
  • Next step: decide which version is official and attach proof.
Explainability engineofficial fact, conflict or gap, likely AI risk
External signalsmentions, discovery gaps, duplicates and monitored sites

Owner confirmations

audit trail

Turn draft facts into owner-backed facts

Some facts should not be guessed by a crawler. The owner confirms the claims AI may safely treat as current: services, regions, contact route, limits and positioning.

  • Why it matters: owner-confirmed facts are stronger than scraped text.
  • Next step: verify ownership, then confirm the important fields.

Ownership verification

Required before owner-confirmed facts become evidence.
unverified

Prove the profile belongs to the business

Verification protects the profile from being changed by the wrong person. It also makes the canonical layer more trustworthy for agencies, clients and AI systems.

  • Methods: domain file, website proof, work email or public profile code.
  • Next step: place the verification token and run the check.
Verification methodsdomain file or public social profile code
Latest attempttoken, placement, server check result

Deployment integrity

technical layer

Check whether the AI-readable layer is actually live

This tab checks public machine-readable surfaces such as `llms.txt`, profile JSON, schema and agent endpoints. It separates “we generated files” from “bots can read them”.

  • For SEO delivery: this is installation QA and proof of work.
  • Next step: publish missing files, then re-check from the server.
Machine-facing deploymentdeployed, synced, warning or missing
Installation checkerserver checks when available, honest fallback otherwise
File outputsgenerated or planned profile surfaces
Brief gapsquestions needed before stronger files are published

AI Bot Intelligence

live

See who read the layer and what it means

Analytics turns server/CDN logs into an operator view: bot visits, blocked paths, failed reads, last bot read, and which endpoints were reached.

  • Not just traffic: it interprets access into risks and next actions.
  • Next step: import logs or connect hosting/CDN data.
Bot behavior interpreterwho came, what it likely was, what it read, what it means
Server log importNginx combined, Cloudflare JSONL, Vercel/hosting JSON logs
Use server/CDN access logs, not browser analytics. IPs are hashed before storage.
Background-ready queuesource scans, file checks, endpoint reads and real conflicts
Background jobsdurable queue foundation for scans and checks
Crawler access matrixsearch, training, user-triggered agents

Recent evidence

Changelog

Freshness

Review dates, stale evidence and recrawl work for the canonical profile.

Keep the AI profile from going stale

AI answers can reuse old facts for months. Freshness tracks what needs re-checking: source reads, owner confirmations, policies, prices and service changes.

  • For retainers: this is the monthly maintenance layer.
  • Next step: queue recrawls and confirm changed facts.
Review queuefreshness evidence used by Trust & Visibility Index

Agent permissions

Draft boundaries for what agents may say or do using the canonical profile.
agent-permissions.json

Define what AI agents are allowed to say or do

Permissions explain safe boundaries: what can be quoted, what needs human review, which facts are public, and which actions should never happen automatically.

  • Why it matters: it reduces risky AI handoffs and overclaiming.
  • Next step: set allowed claims, restricted claims and approval rules.
LLM review queue
Agent action readiness
Evidence bundle

Safe actions

Agent-ready workflows, required data and approval boundaries.
agent-actions.json

Map the next step after AI recommends the business

Actions tell an assistant where to route a user: booking, contact, checkout, quote request, human review or support. The goal is useful handoff, not autonomous risk.

  • For local clients: this exposes the path from answer to lead.
  • Next step: define required fields and confirmation boundaries.

Question coverage

Check whether the current profile and sources can answer real customer questions without forcing AI to guess.
source coverage

Turn customer questions into an AI-readability work queue

Visibility reports show where a brand appears. Question coverage shows whether the business has enough clear, current and source-backed information for agents to answer what buyers actually ask.

  • For agencies: this turns another report into concrete fixes.
  • Next step: add real buying questions, run coverage, then fill the missing facts.
Type your own questions here. Coverage uses canonical facts, source links, conflicts, entity graph, files and prior AI probes.

AI visibility probes

Ask model and answer surfaces what they know, then check official-site presence, citations and canonical accuracy.
source reflection

Test real questions customers may ask AI

Probes compare AI answers with the canonical profile: does AI know the official site, services, contact route, source links and limits?

  • No ranking promise: this is evidence about ambiguity and accuracy.
  • Next step: run a probe, then turn missing facts into profile updates.
Agent simulation
AI surface matrix
Entity position report
Competitor benchmark
Source path attribution
Source fix queue
Checks official site visibility, citation position, canonical match, missing facts and wrong facts.

Repair queue

The action list for improving the AI layer

Client pack tasks combine source conflicts, missing files, bot access, stale facts and AI probes into prioritized work a GEO specialist can send, assign or monitor.

  • For GEO specialists: use this as the client-facing delivery pack and monthly maintenance queue.
  • Next step: fix high-impact tasks first, then verify the result.
Client delivery workspacegenerate, share and update
No saved pack yet.
AI repair systemreason, repair path and verification impact