TOPICALAUTHORITY.ORG TAO / ROOT

AI Search Research Methodology

MTH / 12 · AI RETRIEVAL OBSERVATORYPROMPT → RESPONSE → CLAIM → SOURCE → VARIANCE
TAO METHODS / GENERATIVE RETRIEVAL

AI Search Research Methodology.

An AI-generated answer is not a ranking list. It is a time-bound synthesis produced under a particular prompt, interface state, retrieval context and generation run. Research must capture the complete observation chain—not only the final text.

OBSERVATION / SYNTHETIC TOPICAL QUERYRUN SET ACTIVE
PROMPT FAMILY / TA-01How should a site build defensible topical authority?
AI SEARCH
RESPONSE
CLAIMS
07 OBSERVED
CITATIONS
04 RESOLVED
SOURCES
03 UNIQUE
VARIANCE
02 MATERIAL
PROMPTS05
REPEATS03
CAPTURES15
STATUSILLUSTRATIVE
01 / CONTROLVersion the prompt family
02 / REPEATCapture multiple runs
03 / SEGMENTExtract atomic claims
04 / RESOLVEIdentify cited sources
05 / TESTValidate claim support
06 / BOUNDReport observed variance
MTH / 12.1 DEFINITION

Study the response as an event. Study retrieval as a system.

AI search research observes how a controlled query family produces claims, references and source selections across repeated captures. It does not infer an invisible internal process from one visible answer.

OPERATIONAL DEFINITION / TAO
AI search research is the controlled observation of generated answers, retrieved evidence and run-to-run variation under explicitly recorded query and environment conditions.
OBSERVABLE / RESPONSEText, structure, ordering and visible references

Capture the returned state exactly as observed.

RESOLVABLE / SOURCEDocuments reached through visible citations

Resolve identity, page state and the passage relevant to each claim.

TESTABLE / VARIANCEDifferences across controlled prompts and repeats

Change one dimension at a time and preserve every capture.

UNOBSERVED / INTERNALHidden selection and generation mechanisms

Do not convert output patterns into unsupported internal explanations.

MTH / 12.2 RESEARCH OBJECTS

Eight objects. None can substitute for another.

A domain mention is not a citation. A citation is not proof of claim support. A supported claim in one response is not stable visibility.

OBJECT / 01

Query family

A versioned group of prompts expressing one information need through controlled formulations.

UNIT / PROMPT
OBJECT / 02

Response

The complete visible answer returned for one prompt, environment state and execution.

UNIT / CAPTURE
OBJECT / 03

Atomic claim

One independently testable proposition extracted without changing its original meaning.

UNIT / PROPOSITION
OBJECT / 04

Mention

A named entity, domain, brand, author or concept appearing in the visible response.

UNIT / REFERENCE
OBJECT / 05

Citation link

A visible response element that points to an externally resolvable source object.

UNIT / LINK
OBJECT / 06

Source document

The identified page or document state reached from a citation or retrieval reference.

UNIT / DOCUMENT
OBJECT / 07

Grounding relation

The tested relationship between a response claim and evidence present in its cited source.

UNIT / CLAIM–SOURCE PAIR
OBJECT / 08

Variation event

A material difference in claims, citations, sources or conclusions across comparable runs.

UNIT / DELTA
MTH / 12.3 OBSERVATION PIPELINE

Twelve stages from question to bounded finding.

The pipeline preserves what was asked, what was returned, which evidence was visible and which conclusions survived repeated observation.

01 / QUESTION

Define the information need

Name population, concept, outcome and observation window.

OUTPUT / RQ VERSION
02 / FAMILY

Construct prompt variants

Separate equivalent wording from changed intent or constraints.

OUTPUT / PROMPT MATRIX
03 / CONTEXT

Freeze environment state

Record language, region, date, surface and observable settings.

OUTPUT / RUN CONTEXT
04 / CAPTURE

Execute repeated runs

Preserve complete outputs and timestamps without selective retention.

OUTPUT / RAW RESPONSES
05 / SEGMENT

Extract atomic claims

Split compound statements into independently testable units.

OUTPUT / CLAIM LEDGER
06 / RESOLVE

Map mentions and citations

Distinguish names, links, domains, documents and duplicate sources.

OUTPUT / REFERENCE MAP
07 / SNAPSHOT

Preserve source state

Record the source version inspected for the grounding decision.

OUTPUT / SOURCE REGISTRY
08 / GROUND

Test claim support

Code direct, partial, contradictory, absent or inaccessible support.

OUTPUT / SUPPORT MATRIX
09 / COMPARE

Measure run variation

Compare inclusion, wording, citations, source identity and decision.

OUTPUT / DELTA REGISTER
10 / CLUSTER

Map source concentration

Locate recurring domains, document types and topical subtrees.

OUTPUT / SOURCE GRAPH
11 / LIMIT

Test uncertainty

Separate observable findings from hidden mechanism assumptions.

OUTPUT / CLAIM CEILING
12 / REPORT

Publish bounded results

State scope, run count, measures, exceptions and raw evidence route.

OUTPUT / STUDY RECORD
MTH / 12.4 RESPONSE ANATOMY

One answer contains multiple evidence states.

Response-level scoring hides whether individual claims are supported, partially supported, unsupported or uncited.

CLAIM SEGMENTATION / SYNTHETIC

Break synthesis into testable propositions.

The answer is stored intact, then copied into an atomic claim ledger. Every claim receives a stable identity and a separate relation to each visible citation.

No citation is credited merely because it appears near a paragraph.

CAPTURE / TA-01-P03-R0207 CLAIMS · 04 CITATIONS
C-01. Topical coverage requires an explicit subject boundary.DIRECT SUPPORT
C-02. Internal links can express relationships between supporting pages.PARTIAL SUPPORT
C-03. Entity consistency reduces ambiguity across the content system.UNCITED
C-04. Publishing more pages automatically produces authority.CONTRADICTED
RESOLVED SOURCES03 UNIQUE
SUPPORTED CLAIMS01 DIRECT
MATERIAL CONFLICTS01 FOUND
ILLUSTRATIVE ONLY / The statements and counts demonstrate coding states. They are not observations from a live AI search surface.
MTH / 12.5 VARIANCE CONSOLE

Control one dimension. Observe what changes.

Every panel holds the information need constant while testing a different source of response instability.

CONTROLLED RUN COMPARISON / TOPICAL QUERY FAMILY04 TESTS LOADED
TEST / 01 · SEMANTICALLY ALIGNED PROMPTS

Equivalent intent can still shift source selection.

DIMENSION → WORDING
VARIANT / AHow do sites build topical authority?

Broad explanatory formulation without requested format.

3 SOURCES · 6 CLAIMS
VARIANT / BWhat creates defensible topical authority?

Adds an evidence-quality constraint while preserving the core need.

4 SOURCES · 7 CLAIMS
VARIANT / CExplain topical authority as a research system.

Requests a methodological frame and changes response organization.

3 SOURCES · 8 CLAIMS
MENTION OVERLAPMODERATE
SOURCE OVERLAPPARTIAL
DECISION STABILITYSTABLE
INTERPRETATIONWORDING-SENSITIVE
PERMITTED / Report response patterns across the versioned prompt family. Do not select the variant that produces the preferred answer.
TEST / 02 · SAME PROMPT, THREE CAPTURES

One prompt can produce multiple valid observations.

DIMENSION → EXECUTION
RUN / 01Seven claims; four citations.

Entity coverage and internal architecture appear in the synthesis.

BASELINE CAPTURE
RUN / 02Six claims; three citations.

One architecture claim disappears; two source identities persist.

1 MATERIAL DELTA
RUN / 03Eight claims; four citations.

A new measurement claim appears without direct support.

2 MATERIAL DELTAS
CLAIM PERSISTENCE5 / 8
CITATION PERSISTENCE2 / 5
CORE CONCLUSIONSTABLE
INTERPRETATIONRUN VARIANCE
PERMITTED / Treat each response as one observation in a run set. Visibility is estimated across repeated captures, not declared from the best capture.
TEST / 03 · SOURCE STATE

A citation may resolve while its evidence has changed.

DIMENSION → TIME
CAPTURE STATECitation points to Document D-03.

The response claims that the document defines four coverage layers.

RESPONSE SAVED
INSPECTION STATECurrent document shows three layers.

The accessible page was revised after the response capture.

SOURCE DRIFT
DECISIONGrounding remains unresolved.

Current absence cannot establish what the response-time document contained.

SNAPSHOT REQUIRED
LINK RESOLUTIONSUCCESS
STATE IDENTITYFAILED
SUPPORT DECISIONUNKNOWN
INTERPRETATIONDATE-BOUND
PERMITTED / Separate link availability from historical evidence support. A live page is not automatically the source state used at response time.
TEST / 04 · CLAIM–SOURCE RELATION

Citation presence does not establish evidential support.

DIMENSION → SUPPORT
CLAIM / C-04More pages automatically create authority.

Atomic proposition is causal and unconditional.

HIGH CLAIM STRENGTH
SOURCE / D-02Discusses coverage with quality conditions.

The cited passage does not support automatic causality.

SEMANTIC MISMATCH
DECISION / G-07Contradicted or unsupported.

Code according to the exact relationship defined in the protocol.

REVIEW REQUIRED
CITATIONPRESENT
SOURCERESOLVED
DIRECT SUPPORTABSENT
INTERPRETATIONNOT GROUNDED
WITHHOLD / Do not count the claim as grounded. Citation visibility, link resolution and semantic support are three separate decisions.
MTH / 12.6 PROMPT LATTICE

Prompts are experimental conditions. Version every difference.

The lattice separates stable semantics from deliberate changes in audience, task, format and constraint.

VARIANTINFORMATION NEEDAUDIENCETASKCONSTRAINTFORMATREPEATSCOMPARISON ROLE
P-01 / BASEBuild topical authorityUnspecifiedExplainNoneOpen3BASELINE
P-02 / EVIDENCEBuild topical authorityResearcherExplainUse defensible evidenceOpen3CONSTRAINT
P-03 / PROCESSBuild topical authoritySite operatorDesign stepsName dependenciesOrdered system3TASK
P-04 / CRITIQUEBuild topical authorityAnalystEvaluateInclude failure modesArgument3FRAME
P-05 / DRIFTBuild topic visibilityUnspecifiedExplainNoneOpen3SEMANTIC DRIFT
CONTROL / P-05 is not treated as equivalent because “visibility” changes the target concept.SYNTHETIC MATRIX / 15 PLANNED CAPTURES
MTH / 12.7 CITATION TRACE

Trace every claim from response to source passage.

The trace prevents a nearby citation, repeated domain or reputable source from receiving automatic credit for a claim it does not support.

RESPONSE
R-02Captured answerP-03 · repeat 02 · dated state
CLAIMS
C-01Boundary claimatomic · descriptive
C-04Automatic authority claimatomic · causal
CITATIONS
L-01Visible referenceresolved · D-01
L-03Visible referenceresolved · D-02
SOURCES
D-01Method documentsnapshot identified
D-02Coverage documentcurrent state only
GROUNDING
G-01Direct supportC-01 ↔ D-01
G-07ContradictedC-04 ↔ D-02
TRACE RULE / The unit of grounding is the claim–source pair, not the whole answer and not the source domain.EXAMPLE DATA / SYNTHETIC
MTH / 12.8 MEASURES

Six diagnostics. No universal AI visibility score.

Each measure answers a specific question about appearance, stability, sourcing or evidential support.

MEASURE / 01

Response inclusion rate

responses containing target object ÷ eligible responses

Measures whether a defined entity, source or concept appears in the captured run set.

VISIBILITY / RESPONSE LEVEL
MEASURE / 02

Claim persistence

runs containing equivalent claim ÷ comparable runs

Tests whether a proposition survives repeated execution under the same condition.

STABILITY / CLAIM LEVEL
MEASURE / 03

Citation persistence

runs citing resolved source ÷ comparable runs

Separates stable source selection from one-off citation appearance.

STABILITY / SOURCE LEVEL
MEASURE / 04

Direct support rate

directly supported cited claims ÷ tested cited claims

Measures claim–source grounding after citation identity and source state are resolved.

EVIDENCE / RELATION LEVEL
MEASURE / 05

Source concentration

citations from leading source set ÷ resolved citations

Shows whether observed sourcing is distributed or repeatedly dependent on a narrow set.

DIVERSITY / CORPUS LEVEL
MEASURE / 06

Material variance rate

decision-relevant deltas ÷ compared response elements

Counts only changes capable of altering the reported finding or claim boundary.

UNCERTAINTY / DECISION LEVEL
MTH / 12.9 WORKED EXPERIMENT

Five prompts. Three repeats. One bounded interpretation.

The example demonstrates how topical concepts, citations and support states are compared without presenting synthetic values as live findings.

AIR
SYNTHETIC CASE / TOPICAL AUTHORITY QUERY FAMILY15 response captures · 5 prompt variants · 3 repeats per variant
ILLUSTRATIVE
QUESTIONWhich concepts persist?Boundary, entity coverage, internal relationships, evidence and measurement.
OBSERVATIONThree concepts recur broadly.Two appear only under methodological or critique framing.
SOURCESSeven unique documents.Two source identities recur across more than one prompt condition.
GROUNDINGSupport varies by claim.Some citations support only part of the surrounding synthesis.
BOUNDARYNo market-wide claim.The result describes this prompt family, capture window and observed surface.
CONCEPT / RUN
P-01
P-02
P-03
P-04
P-05
SUBJECT BOUNDARY
3 / 3
3 / 3
3 / 3
3 / 3
2 / 3
ENTITY COVERAGE
2 / 3
3 / 3
3 / 3
2 / 3
1 / 3
INTERNAL RELATIONS
3 / 3
2 / 3
3 / 3
3 / 3
2 / 3
EVIDENCE QUALITY
1 / 3
3 / 3
2 / 3
3 / 3
1 / 3
INTERPRETATION / Boundary and internal relations are more persistent than evidence language in this synthetic run set.NOT A LIVE AI SEARCH CLAIM
MTH / 12.10 OBSERVATION MODEL

AI answers and ranked results require different units.

The methods can share query controls and source validation, but their observable structures are not interchangeable.

DIMENSION
RANKED RESULT OBSERVATION
AI RESPONSE OBSERVATION
Primary unit
Result object at an observed position.
Generated response containing claims and references.
Ordering
Explicit visible sequence of results.
Narrative placement, emphasis and synthesis order.
Source relation
Result links directly to one destination.
Claim may relate to zero, one or several citations.
Repeat variance
Positions and result composition can change.
Claims, wording, sources and citations can all change.
Support test
Inspect whether destination answers the query.
Test every atomic claim against every attributed source.
Permitted claim
Describes observed result presence and position.
Describes response inclusion, persistence and grounding.
MTH / 12.11 FAILURE MODES

Eight shortcuts that create false AI search findings.

These errors collapse distinct objects, select convenient runs or claim access to mechanisms that were never observed.

FAILURE / 01

Single-run certainty

One response is presented as stable visibility across time and prompt conditions.

FIX / REPEAT CAPTURES
FAILURE / 02

Prompt cherry-picking

Only the wording that produces the desired inclusion is retained.

FIX / VERSION FAMILY
FAILURE / 03

Mention–citation collapse

A named entity is counted as cited even when no source link attributes it.

FIX / SEPARATE OBJECTS
FAILURE / 04

Citation–support collapse

A visible citation receives credit without testing the cited passage.

FIX / TEST CLAIM PAIR
FAILURE / 05

Answer-level grounding

Mixed support states are hidden inside one score for the whole response.

FIX / ATOMIC CLAIMS
FAILURE / 06

Live-source substitution

A current source page silently replaces the state available at capture time.

FIX / PRESERVE STATE
FAILURE / 07

Mechanism invention

Output patterns are described as proof of hidden retrieval or ranking logic.

FIX / BOUND INFERENCE
FAILURE / 08

Score compression

Visibility, persistence, citation and support are collapsed into one opaque number.

FIX / REPORT MEASURES
MTH / 12.12 STUDY RECORD

Every finding must retain its observation coordinates.

The study record makes each response, claim, citation, source and comparison independently traceable.

AI SEARCH STUDY RECORD / MTH12-V1
01 / QUESTIONVersioned research targetpopulation + concept + outcome + time
02 / PROMPTSComplete query familyvariant + semantics + role + version
03 / CONTEXTObservable run conditionssurface + language + region + date
04 / CAPTURESRaw response registryrun ID + timestamp + preserved output
05 / CLAIMSAtomic proposition ledgerclaim ID + wording + strength + state
06 / REFERENCESMention and citation mapidentity + placement + resolution
07 / SOURCESDocument state registryURL + identity + snapshot + access
08 / GROUNDINGClaim–source decisionsdirect + partial + absent + conflict
09 / VARIANCEMaterial delta registerprompt + run + source + claim change
10 / MEASURESDeclared calculationsunit + denominator + exclusions
11 / LIMITSClaim boundary recordunknowns + sensitivity + exclusions
12 / FINDINGBounded research statementresult + scope + evidence + version
MTH / 12 · OBSERVATION PRINCIPLE

Do not measure an answer. Measure the system of observations.

AI search research becomes defensible when prompts are controlled, responses are repeated, claims are segmented, citations are resolved, sources are inspected and every conclusion remains bounded by visible evidence.

TOPICALAUTHORITY.ORG / METHODS / MTH-12CONTROL → CAPTURE → SEGMENT → RESOLVE → GROUND → COMPARE → BOUND
TAO / CONTACT · DIRECT TRANSMISSION Have an asset, domain or market position to investigate? ENTER CONTACT SYSTEM →
DIGITAL ASSET INTELLIGENCE + EXECUTION
EXECUTED BY
BB DIGITALNA AGENCIJA

Investigation, consulting and execution of digital assets, premium-domain strategies, information architecture, semantic systems, websites and agreed digital growth plans.

TOPICALAUTHORITY.ORG / SEMANTIC INTELLIGENCE SYSTEM BB DIGITALNA AGENCIJA / BB.HR