INFRASTRUCTURE & DISTRIBUTION · AGENTIC ANALYTICS

AI Search Analytics From Recommendations to Revenue.

Run governed AI search monitoring across a defined buyer-question cohort. Lumenario combines AI visibility tracking, citations, recommendation share, a scoped visibility score, and attributed traffic with finalized Google Search Console, crawler, and owned-site evidence.

See What We Measure See the ROI Evidence Model
Versioned Question CohortsGSC + Site SignalsDisclosed Evidence Boundaries
REPRESENTATIVE PRODUCT WORKSPACE 01 / 03
Representative Agentic Analytics dashboard showing anonymized crawl hits, successful fetches, and response health.
01 · CRAWLER OVERVIEWCrawler health at a glance.Track total requests, successful fetches, and non-2xx responses across the selected period.
Representative day-to-day crawler activity and anonymized counts for Googlebot, Bingbot, OAI SearchBot, and ClaudeBot.
02 · ACTIVITY BY BOTDaily request activity and bot mix.Compare request movement and volume for each crawler observed in the workspace.
Representative Agentic Analytics table showing anonymized crawler requests, delivery success, and activity by page.
03 · PAGE-LEVEL ACTIVITYSee which pages crawlers request.Review page-level request volume, delivery results, and last-seen activity.
ONE CONNECTED MEASUREMENT SYSTEM

See the answer, the search response, and the owned-site signals.

Each layer records a different kind of evidence. Agentic Analytics keeps the signals separate, then joins them through shared page URLs and content identifiers where a reliable match exists.

01 · AI ANSWERS

What do answer systems say?

Repeated observations show whether the brand is retrieved when provider evidence is available, mentioned, cited, considered, or actively recommended.

  • Prompt and buyer intent
  • Provider and model
  • Region and cohort
  • Competitor role and evidence
FINEST VIEW · INDIVIDUAL ANSWER RUN
02 · GOOGLE SEARCH

How is organic discovery moving?

Google Search Console adds organic-search response data for mapped pages and topics.

  • Impressions and clicks
  • CTR and weighted position
  • Page and query
  • Country, device, appearance
FINEST VIEW · DATE × PAGE × QUERY
03 · OWNED SITE

What happens on the destination?

Edge logs record successful public page requests. A separate first-party beacon records rendered views and link selections.

  • Page requests and rendered views
  • AI, search, direct, and referral sources
  • Per-tab sessions
  • Internal, outbound, and CTA clicks
FINEST VIEW · HOUR OR EVENT × PAGE × SERVED VERSION
04 · CRAWLER REQUESTS

Which crawlers request the public knowledge?

Crawler fetches show request and delivery activity. They do not prove indexing, model training, citation, or answer use.

  • Provider, agent, and purpose
  • Verification status
  • Page, resource, and hour
  • Status, cache, bytes, and duration
FINEST VIEW · HOUR × PAGE × SYSTEM
RECOMMENDATION SHARE

Measure the questions your buyers ask—not an imaginary global score.

The monitored universe is agreed before reporting begins: buyer questions, AI systems, models, markets, languages, competitor set, sampling frequency, and eligibility rules.

01 · COMPETITIVE CONTEXT

Recommendation Share

Of the observed recommendations for brands in your locked comparison set, what portion belonged to your brand?

HOW MUCH OF THE TRACKED CATEGORY DO WE EARN?
02 · ABSOLUTE FREQUENCY

Recommendation Rate

Across every eligible answer run in the monitored cohort, how often was your brand actually recommended?

HOW OFTEN ARE WE RECOMMENDED AT ALL?

Scope matters. Both numbers use the same versioned questions, systems, models, markets, and competitors. Share excludes untracked brands, is unavailable when no tracked recommendations occur, and is never presented as global market share.

  1. 01 · DEFINE

    Lock the universe.

    Version the questions, providers, models, markets, competitors, and baseline.

    OUTPUT · MEASUREMENT CONTRACT
  2. 02 · OBSERVE

    Run repeatedly.

    Store timestamped answers and sources because generative output varies.

    OUTPUT · ANSWER RUNS
  3. 03 · CLASSIFY

    Separate each role.

    Distinguish mentions, retrieved evidence, citations, and recommendations.

    OUTPUT · AUDITABLE EVENTS
  4. 04 · COMPARE

    Measure movement.

    Compare matched cohorts, disclose run counts, and retain variability.

    OUTPUT · SHARE + RATE
CONTENT-LEVEL TRACEABILITY

Join evidence to the correct content identity—without inventing a causal path.

AI answers, GSC, crawler requests, and site events are parallel evidence layers. Canonical URLs and publication keys connect them where possible; served versions remain available for first-party and edge events.

AI ANSWERSPrompt · provider · model · runSource or citation URL when present
GOOGLE SEARCHDate · page · query · marketMapped canonical page
SHARED KEYCanonical URL
+ publication identityVersion retained for site and edge events
OWNED SITEEvent · page · served versionReferrer, session, and clicked element
CRAWLERSHour · page · system · purposeStatus, cache, bytes, and duration
Example Agentic Analytics metric and evidence ledger
SignalFinest dimensionLinked objectEvidence classWhat it proves
AI recommendationPrompt × provider × model × runAnswer + cited URL when presentObservedBrand was recommended in this answer
Google Search performanceDate × page × queryMapped canonical pageObservedSearch visibility and response moved
Recognized AI-referral rendered viewEvent × page × versionServed publication versionConnectedThe tracker rendered a page with a recognized AI referrer or UTM
Crawler fetchHour × page × systemPublic URL / publicationOperationalRequest and delivery activity only
Commercial contributionConnected journey × outcomeCRM/conversion recordAttributedValue under the agreed attribution rule
A DEFENSIBLE ROI MODEL

Prove value at the level the evidence allows.

Observed AI, search, and site movement is available by default. Financial ROI is modeled only after business outcomes, margin, savings, costs, and an attribution rule are connected and approved.

01 · OBSERVED PERFORMANCE

What the systems recorded

  • Recommendation, mention, and citation events
  • GSC impressions, clicks, CTR, and position
  • Site visits, referrers, and link selections
EVIDENCE · DIRECTLY OBSERVED
02 · CONNECTED OUTCOMES

What can be attributed under agreed rules

  • Governed conversions or CRM outcomes
  • Defined attribution window and journey
  • Gross profit rather than headline pipeline
EVIDENCE · CONNECTED + ATTRIBUTED
03 · ECONOMIC VALUE

What remains an approved estimate

  • Documented hours actually displaced
  • Agreed loaded cost or external spend avoided
  • All incremental programme costs included
EVIDENCE · MODELED + DISCLOSED
GOVERNED MEASUREMENT RUNS

One operating loop from observation to the next decision.

On each governed run, observations retain their question, system, page, source, and reporting context before Feedback prioritizes hypotheses and governed next actions.

  1. MEASURE AI

    Run the governed question panel.

    Capture answer, role, sources, citations, provider, model, region, and execution time.

  2. JOIN SEARCH

    Bring in GSC response data.

    Connect page, query, market, device, impressions, clicks, CTR, and position.

  3. JOIN SITE

    Observe the owned destination.

    Compare edge page requests with separate rendered-view, acquisition, session, and link-click events.

  4. RESOLVE IDENTITY

    Match shared pages where evidence allows.

    Retain the served version for edge and browser events; map AI citations and GSC facts when their URL resolves.

  5. COMPARE

    Normalize against the baseline.

    Use matched cohorts, explicit denominators, and disclosed uncertainty.

  6. ACT

    Route evidence into Feedback.

    Diagnose what moved and prepare the next governed change.

AGENTIC ANALYTICSMeasures what moved, where, and with what confidence
FEEDBACK AGENTPrioritizes hypotheses and governed next actions
See the Feedback loop →
MAKE AI DISCOVERY ANSWERABLE

Know your share. Show the evidence. Improve the system.

Start with a governed buyer-question universe and a baseline. Lumenario brings AI-answer observations, Google Search performance, crawler requests, and owned-site behaviour into one evidence view, resolving shared page identity where possible.

Book a Measurement ReviewFree AI Visibility Report

Observe. Connect. Attribute. Improve.