System: MWMS
Document Type: Standard
Status: Active
Version: v1.1
Authority: HeadOffice
Applies To: Research Brain, Affiliate Brain, Ads Brain, Content Brain, HeadOffice Intelligence, Data Brain, Experimentation Brain, Risk Brain, AIBS Brain, AI Employees, Deep Search Workflows, Client Intelligence Systems, Future Client Facing AI Systems
Primary Location: MCR
Future Operational Destination: mwmsbrain.site, mwmsheadofficebrain.site, Brain Room, Research Brain, Affiliate Brain, Ads Brain, AI Employee Dashboards, HeadOffice Dashboards, Future Client Portals
Parent Page: HeadOffice
Source Of Truth: MCR
Related Frameworks: MWMS Deep Search Quality And Observability Framework, MWMS Research Planning And Query Rewriting Standard, MWMS AI Guardrail And Preflight Check Standard, MWMS AI Work Session Persistence Standard, MWMS AI Observability Metadata Standard, MWMS AI Employee Evaluation Scorecard Standard, MWMS Agent Loop Control Framework, MWMS Independent Model Review And Rescue Routing Framework, MWMS Research Synthesis Documentation And Distribution Framework, MWMS External Knowledge Engine And Reasoning Agent Separation Framework
Course Source: Matt Pocock AIhero Build DeepSearch In TypeScript and AI Automations by Jack Fact Checking, Multi-Model Research, Browser Copilot, Client Intelligence, Source Verification, and Evidence Display Material
Absorption Status: Approved For Integration
Last Reviewed: 2026-06-21
Purpose
The purpose of this standard is to define how MWMS displays sources, claims, evidence, evidence roles, source confidence, source independence, factual verdicts, and unresolved disagreement to human operators.
MWMS AI Employees must not only produce answers.
They must show:
what claim was evaluated
what evidence was used
where the evidence came from
what each source actually contributed
whether the source supports or challenges the claim
how reliable the source appears
how fresh the source is
whether apparently separate sources rely on the same original evidence
whether evidence is sufficient
whether credible disagreement remains
whether the final conclusion is properly supported
This standard ensures source evidence is not hidden inside backend traces, raw logs, model reasoning, unstructured source lists, or polished reports.
The operator should be able to determine:
which sources were discovered
which sources were inspected
which sources were used
which sources were rejected
which sources were inaccessible
which sources were duplicates or derivatives
what each source supported
what each source challenged
whether the original source was found
whether the claim was verified
whether the conclusion contains interpretation
whether uncertainty remains
The goal is to make AI-generated intelligence easier to trust, review, audit, challenge, correct, and improve.
Scope
This standard applies to any MWMS workflow where sources, evidence, research, client information, or external information influence an output or decision.
This includes:
Deep Search workflows
Research Brain investigations
fact-checking systems
browser copilots
multi-model research
Affiliate Brain offer evaluations
Ads Brain compliance reviews
Content Brain research workflows
HeadOffice Intelligence reviews
Newsletter Intelligence follow-up research
client intelligence reports
meeting preparation research
product or vendor research
tool evaluations
source inspections
market trend research
compliance and policy checks
Experimentation Brain analysis
Risk Brain reviews
AIBS client reports
future client-facing research outputs
future conversational research systems
future evidence-backed dashboards
This standard does not define exact frontend code, React components, scraping implementation, database implementation, or source table structure.
It defines the MWMS governance standard for showing evidence to operators and approved recipients.
Core Rule
The core rule is:
Sources and evidence used by AI Employees must be visible, understandable, and connected to the claim or conclusion they support.
If an AI Employee uses sources to support an answer, the operator should be able to inspect the evidence behind that answer.
A source-backed answer without visible source evidence is incomplete.
A list of links without claim-level connection is also incomplete.
Definition Of Source Visibility
Source visibility means the human operator can see the source evidence behind an AI output.
Source visibility should answer:
What sources did the AI find?
What sources did the AI inspect?
Which original sources were located?
Which sources were derivative?
What did each source say?
What claim did each source relate to?
Did each source support, challenge, qualify, or merely contextualise the claim?
Was the source official, primary, expert, independent, commercial, user-generated, or unknown?
Was the source current enough?
Was the source trusted?
Was the source actually used in the final answer?
Did any source conflict with another?
Did several sources rely on one underlying report?
Did the AI answer rely on strong or weak evidence?
Source visibility turns AI output from a black-box answer into a reviewable intelligence record.
Definition Of Evidence Display
Evidence display is the user-facing presentation of:
the evaluated claim
source material
evidence summaries
evidence roles
trust indicators
freshness indicators
source independence indicators
support and challenge status
limitations
conflicts
evidence sufficiency
verdict
confidence
unresolved issues
Evidence display may appear as:
claim cards
source cards
evidence panels
expandable source lists
citation rows
source summaries
trust and freshness badges
support and challenge labels
source relationship warnings
source conflict warnings
evidence sufficiency notes
claim-to-evidence maps
final answer support maps
HeadOffice review panels
Evidence display is not the same as raw logs.
It should be readable, structured, and useful to the operator.
Why MWMS Needs This Standard
Without source visibility, MWMS risks:
trusting polished but unsupported answers
missing outdated sources
missing weak evidence
hiding source conflicts
counting duplicated reports as independent evidence
treating citations as proof
making poor affiliate decisions
making risky ad compliance decisions
approving weak content
repeating research
losing auditability
weakening HeadOffice oversight
making AI Employees harder to evaluate
making client-facing systems less trustworthy
allowing model interpretation to appear as source fact
presenting confidence as truth
With source visibility, MWMS gains:
operator trust
better review quality
stronger evidence discipline
better compliance safety
better Research Brain outputs
better Affiliate Brain offer decisions
better Ads Brain policy review
better HeadOffice intelligence
reusable source records
claim-to-evidence traceability
clearer disagreement handling
clearer Kaizen learning
Source Visibility Versus Observability
Observability is for tracing what the system did.
Source visibility is for showing the operator what evidence supports the answer.
Observability may show:
model calls
tool activity
database activity
cost
latency
retries
errors
workflow state
Source visibility should show:
claim evaluated
source evidence
evidence role
source quality
freshness
independence
limitations
conflicts
verdict support
Both are needed.
A backend trace may show that a source was inspected.
Source visibility shows whether that source was relevant, useful, current, independent, trusted, and capable of supporting the conclusion.
Source Visibility Versus Citations
Citations show where information came from.
Source visibility shows:
what the source contributed
whether the source supports the exact claim
whether the source is reliable
whether the source is current
whether the source is independent
whether the source has limitations
whether contradictory evidence exists
A citation alone is not enough.
MWMS should aim to show:
source title
source URL or record reference
source type
source owner
publication date
event date where relevant
retrieved date
trust rating
freshness rating
relevance rating
evidence role
evidence summary
claim supported or challenged
source independence
used in final answer status
limitations
conflicts
A link without source context is weak evidence display.
Claim Visibility Standard
Where evidence is used to evaluate a material claim, the claim itself must be visible.
The display should include:
Claim ID:
Original Claim:
Normalised Claim:
Claim Type:
Claimant Or Origin:
Originating Source:
Claim Date:
Time Sensitivity:
Verdict:
Confidence:
Evidence Sufficiency:
Rule
The operator must be able to see what exact claim the evidence was used to evaluate.
The system must not present evidence against a simplified or altered version of the original claim without disclosure.
Claim Types
Possible claim types include:
factual claim
numerical claim
current-status claim
historical claim
causal claim
comparative claim
predictive claim
performance claim
legal or regulatory claim
medical or safety claim
marketing claim
opinion presented as fact
Different claim types may require different evidence standards.
Required Source Display Fields
Each displayed source should include, where available:
Source ID:
Source Title:
Source URL Or Record Reference:
Source Type:
Source Owner:
Original Source ID:
Derived From Source ID:
Publication Date:
Event Date:
Last Updated Date:
Retrieved At:
Freshness Rating:
Trust Rating:
Relevance Rating:
Evidence Role:
Evidence Summary:
Key Claims:
Claim IDs:
Limitations:
Independent Evidence:
Shared Evidence Base:
Used In Final Output:
Conflict Detected:
Source Status:
Verification Status:
Recommended Source Card Format
A source card should be readable at a glance.
Recommended format:
Source Title:
[Title]
Source Type:
Official / Primary Evidence / Expert / Independent Analysis / Commercial / User-Generated / Unknown
Freshness:
Current / Acceptable / Possibly Outdated / Outdated / Unknown / Not Applicable
Trust:
High / Medium / Low / Unknown
Evidence Role:
Supports / Challenges / Partially Supports / Context / Original Data / Repeated Source / Expert Interpretation / Signal Only
Evidence Summary:
[Short summary of what this source contributed]
Related Claim:
[Claim ID or claim summary]
Independent Evidence:
Yes / No / Unclear
Used In Final Answer:
Yes / No / Background Only / Rejected
Limitations:
[What this source does not prove or where caution is needed]
Verification Status:
Verified / Partially Verified / Unverified / Failed / Blocked
This may later become a UI component on mwmsbrain.site or mwmsheadofficebrain.site.
Source Type Classification
MWMS should classify source types.
Official Source
Examples:
vendor documentation
official product page
platform policy page
government page
company announcement
official changelog
affiliate network page
Default trust level:
High for what the organisation officially states, but still checked for bias, scope, and freshness.
Primary Evidence Source
Examples:
research paper
court record
regulatory filing
financial filing
official dataset
direct transcript
technical specification
original survey
original report
Default trust level:
High where authentic, relevant, current, and correctly interpreted.
Expert Source
Examples:
respected industry publication
specialist blog
analyst report
credible technical article
recognised practitioner source
Default trust level:
Medium to high.
Independent Analysis Source
Examples:
reputable journalism
independent investigation
research organisation
credible comparison using disclosed methods
Default trust level:
Medium to high depending on its evidence.
Commercial Source
Examples:
sales page
affiliate review page
comparison page with commissions
agency landing page
vendor marketing page
Default trust level:
Medium to low unless corroborated.
User-Generated Source
Examples:
forums
comments
social media posts
community discussion
Default trust level:
Signal only, not proof.
Unknown Source
Examples:
content farms
anonymous blogs
scraped reposts
pages with unclear ownership
thin AI-generated content
Default trust level:
Low.
Freshness Display Standard
Every source used in time-sensitive workflows should show freshness.
Freshness Ratings
Current
Suitable for current decision-making.
Acceptable
Recent enough for the task.
Possibly Outdated
Use with caution.
Outdated
Should not be trusted for a current decision without historical context.
Unknown
No clear date or freshness signal.
Not Applicable
Evergreen or historical source where current freshness is not material.
Freshness matters most for:
pricing
affiliate payouts
product availability
platform policies
laws and regulations
AI tool features
software documentation
advertising rules
market trends
current events
company leadership
compliance matters
financial performance
Rule
If freshness matters and the source date is unknown, the source should not be displayed as high confidence.
Publication Date And Event Date Rule
The publication date and the date of the event or evidence should remain separate.
A newly published article may discuss old evidence.
An older source may remain the original evidence for a historical event.
The display should show both dates where relevant.
Trust Display Standard
Sources should display a trust rating where possible.
Trust Ratings
High
Official, primary, credible, directly relevant, and sufficiently reliable.
Medium
Useful but may require corroboration or qualification.
Low
Weak, biased, indirect, outdated, commercially conflicted, or questionable.
Unknown
Not enough information is available to judge.
Trust rating should consider:
source owner
source type
publication quality
bias
commercial incentive
directness of evidence
freshness
method transparency
corroboration
history of reliability
Rule
A low-trust source may be shown as a signal, but it should not be treated as proof.
Relevance Display Standard
A source should show how relevant it is to the task and claim.
Relevance Ratings
High
Directly supports, challenges, or answers the material claim.
Medium
Supports part of the claim or provides meaningful context.
Low
Background only or weakly related.
A source can be trustworthy but not relevant.
A source can be relevant but not trustworthy.
Both must be visible.
Evidence Role Standard
Every source used in a material conclusion should be assigned an evidence role.
Possible evidence roles include:
Supports Claim
Challenges Claim
Partially Supports Claim
Qualifies Claim
Provides Context
Defines Terminology
Supplies Original Data
Repeats Another Source
Provides Expert Interpretation
Provides User Experience Signal
Does Not Materially Support Conclusion
Rule
A citation is not automatically supporting evidence.
The operator must be able to see the actual role the source played.
Claim-To-Evidence Mapping Standard
Material claims should be connected to the sources and evidence used to evaluate them.
A claim-to-evidence display should include:
Claim:
Verdict:
Supporting Sources:
Challenging Sources:
Context Sources:
Original Evidence:
Missing Evidence:
Evidence Sufficiency:
Confidence:
Unresolved Issues:
Rule
The operator must be able to determine which source supports which claim.
A general source list at the end of a report is not sufficient for material decisions.
Evidence Summary Rule
Each source should include a short evidence summary.
The summary should answer:
What did this source contribute?
What claim did it support, challenge, or qualify?
Why was it useful?
Was it the original source?
What limitation should the operator know?
The summary must not invent missing evidence.
If the source does not answer the needed question, the summary should say so.
Evidence Summary Versus Source Description
A source description explains what the page or document is.
An evidence summary explains what material evidence it contributes.
Example Source Description:
A company announcement discussing a new product.
Example Evidence Summary:
The company officially announced the product on the stated date but did not provide independent performance evidence.
Rule
Source descriptions must not be mistaken for evidence.
Used In Final Output Rule
MWMS should show whether a source was actually used in the final answer.
Possible statuses include:
Used
Source materially supported or challenged the final conclusion.
Not Used
Source was inspected but not used.
Background Only
Source helped provide context but did not support the final claim.
Rejected
Source was weak, irrelevant, duplicated, outdated, or unreliable.
Failed
Source could not be inspected.
Conflicting
Source materially conflicted with other evidence.
Repeated Source
Source repeated another source without adding independent evidence.
This helps operators understand the evidence path.
Source Independence Standard
Several sources do not necessarily provide several independent confirmations.
Sources may rely on:
the same press release
the same research paper
the same company statement
the same unnamed source
the same data provider
the same syndicated article
the same social media post
the same AI-generated summary
The source display should show:
Original Source:
Derived Sources:
Syndicated Content:
Shared Evidence Base:
Independent Reporting:
Independent Evidence:
Rule
Ten pages repeating one report are one evidence chain, not ten independent confirmations.
Source Independence Display
Possible independence statuses include:
Independent
The source provides materially separate evidence or reporting.
Dependent
The source relies on an identified original source.
Shared Evidence Base
The source appears separate but relies on the same underlying evidence.
Unclear
Source dependence cannot be determined.
Not Applicable
Independence is not relevant to the use of the source.
Rule
The number of source URLs must not be used as a substitute for evidence diversity.
Original Source Preservation Standard
Where practical, the original source should be preserved and displayed alongside derivative reporting.
The system should identify:
original claim source
original evidence source
later reporting
commentary
reposts
summaries
AI-generated interpretation
Rule
MWMS should not use a derivative article as the only visible evidence where the original source is accessible.
Original Source And Verification Separation
The original source shows what was originally claimed or published.
Independent verification determines whether the claim is supported.
The display should distinguish:
Original Claim Source
Independent Verification Source
Contradictory Source
Context Source
Expert Interpretation
Rule
A company’s own statement is evidence that the company made the statement.
It is not automatically independent evidence that the statement is true.
Source Conflict Display Rule
If sources conflict, the operator should see it.
Conflict display should include:
conflicting sources
claim affected
what they disagree on
source dates
source types
which evidence appears stronger
whether different definitions or time periods explain the disagreement
whether the conflict affects the verdict
whether human review is required
Rule
Source conflict should lower confidence unless resolved.
The AI Employee must not hide disagreement between sources.
Unresolved Disagreement Display Standard
Where credible disagreement remains unresolved, the output should display:
Unresolved Issue:
Position One:
Evidence Supporting Position One:
Position Two:
Evidence Supporting Position Two:
Reason Unresolved:
Risk If Position One Is Wrong:
Risk If Position Two Is Wrong:
Required Next Step:
Execution Restriction:
Human Review Status:
Rule
Unresolved disagreement must not be converted into false consensus.
Verdict Display Standard
Where the workflow evaluates a factual claim, the output should use a controlled verdict.
Recommended verdicts include:
Supported
Strong available evidence supports the claim.
Mostly Supported
The central claim is supported, but qualifications or minor inaccuracies exist.
Partially Supported
Some parts are supported, but the complete claim is not.
Misleading
The claim uses true elements but creates a materially incorrect impression.
Unsupported
Sufficient supporting evidence was not found.
Contradicted
Strong evidence conflicts with the claim.
Outdated
The claim may previously have been accurate but is no longer current.
Unverifiable
Available evidence is insufficient to reach a reliable conclusion.
Opinion Or Prediction
The statement cannot be treated as a settled factual claim.
Rule
The system should not force every claim into true or false.
Verdict And Explanation Separation
The verdict and the explanation must remain separate.
The verdict provides the controlled classification.
The explanation should describe:
what evidence supports the verdict
what limitations exist
what evidence challenges it
what uncertainty remains
Rule
A confident explanation must not silently strengthen a cautious verdict.
Confidence Display Standard
Confidence must remain separate from the verdict.
Verdict answers:
What does the available evidence indicate?
Confidence answers:
How strongly can MWMS rely on that verdict?
A claim may be:
Supported with moderate confidence
Unsupported with low confidence
Contradicted with high confidence
Unverifiable with high confidence that evidence is insufficient
Rule
Confidence is not truth.
Confidence must reflect evidence quality, source independence, freshness, coverage, and unresolved disagreement.
Evidence Sufficiency Display
The final output should include an evidence sufficiency rating where relevant.
Strong
Evidence is reliable, current, sufficiently independent, and directly supports the conclusion.
Acceptable
Evidence supports the conclusion but has minor limitations.
Weak
Evidence is incomplete, indirect, dependent, old, or uncertain.
Insufficient
There is not enough evidence to support a trusted conclusion.
Evidence sufficiency should influence:
confidence
human review requirements
final recommendation
whether the workflow continues
whether the result is parked
whether execution is restricted
Rule
Evidence sufficiency must not be determined by source count alone.
Source-List Theatre Prevention Rule
MWMS must prevent source-list theatre.
Source-list theatre occurs when an output appears well researched because it contains many links, citations, or source cards, but the displayed sources:
do not support the claim
repeat the same original evidence
are not inspected
are outdated
are irrelevant
are low trust
contain only search snippets
are included only to make the report appear authoritative
A source display must show:
whether the source was inspected
what evidence it contributed
which claim it relates to
whether it is independent
whether it was used
what limitation applies
Rule
A long source list is not proof of strong research.
Search Snippet Display Rule
Search snippets may help source discovery.
They should not normally be displayed as full evidence.
Where a snippet is used because the source cannot be opened, the display should state:
Source Inspection Status: Failed Or Blocked
Evidence Type: Search Snippet Only
Confidence Impact:
Verification Required:
Rule
Search snippets must not be presented as equivalent to inspected source content.
Source Display By Workflow Type
Different workflows require different source display depth.
Deep Search
Should show:
evaluated claims
all inspected sources
evidence summaries
evidence roles
trust and freshness ratings
source independence
used-in-answer status
conflicts
verdicts
confidence
evidence sufficiency
Affiliate Offer Evaluation
Should show:
vendor or product source
affiliate network source
terms and payout source
customer complaint or review signals
competitor sources
compliance or policy sources
source independence
current availability
Ads Compliance Review
Should show:
official policy source
product claim source
jurisdiction source
compliance risk evidence
policy date
source freshness
claim-to-policy relationship
Content Research
Should show:
topic sources
primary authority sources
competing content
freshness
search intent evidence
original evidence
claim suitability
HeadOffice Intelligence
Should show:
original newsletter or source
supporting external sources
challenging evidence
business relevance evidence
routing basis
risk notes
confidence
Client Intelligence
Should show:
client-approved source
external research
source ownership
privacy classification
verified client facts
interpretation
recommendation basis
Client-Facing Output
Should show sources carefully, with privacy and simplicity controls.
Client-facing evidence may be simplified, but the internal evidence record must remain available.
Multi-Model Research
Should show:
which model produced each interpretation
which sources each model used
whether source sets were independent
areas of agreement
areas of disagreement
minority evidence
final synthesis basis
Rule
Model agreement must not be displayed as source agreement unless the evidence is independently supported.
Operator UI Source Panel
Future MWMS interfaces should include an operator source and evidence panel where useful.
A source and evidence panel may include:
claim list
source list
source cards
evidence cards
filters by source type
filters by trust
filters by freshness
filters by evidence role
filters by independence
used versus unused sources
original versus derivative sources
conflicts
unresolved disagreements
evidence summary
source inspection status
claim-to-evidence map
open source link
copy source summary
create source record
route source issue to Research Brain
mark source as weak
mark source as derivative
add human review note
This is especially valuable for HeadOffice, Research Brain, Affiliate Brain, Ads Brain, and AIBS Brain.
Source Visibility Levels
Not all users need the same source visibility.
Internal Full Evidence
Full source, claim, evidence, independence, and disagreement detail for authorised MWMS operators.
HeadOffice Review
Evidence plus risk, confidence, verdict, independence, and decision indicators.
Developer Debug
Source plus retrieval, tool, database, and failure metadata.
Client Summary
Simplified sources, explanation, confidence, and limitations.
Hidden Or Internal Only
Sensitive sources not shown externally.
Archived Evidence
Historical source and evidence records preserved for audit.
Rule
Internal MWMS evidence should be richer than client-facing evidence.
Source Privacy And Safety Rule
Source display must not expose sensitive information unnecessarily.
Do not expose:
private credentials
private user data
restricted client data
internal API responses
private financial details
confidential source excerpts
copyrighted long-form content beyond what is permitted
private email content unless authorised
restricted employee information
Where needed, display:
source reference
short summary
redacted excerpt
internal-only tag
access-restricted link
privacy classification
Evidence Display In AI Work Sessions
Source visibility should connect to the MWMS AI Work Session Persistence Standard.
A persistent AI work session should store:
claim IDs
source IDs
source summaries
source ratings
evidence roles
source independence
evidence sufficiency
sources used in final answer
conflicts detected
unresolved disagreement
verdict
confidence
review notes
source-related Kaizen notes
This makes source work reusable across sessions.
Evidence Display In Agent Loop Context
Source visibility should connect to the MWMS Agent Loop Context Schema.
Agent loop context should track:
claims
selected sources
inspected sources
source summaries
evidence items
evidence roles
evidence sufficiency
source conflicts
source dependence
answer readiness
review requirement
This helps the Next Action Picker decide whether to:
continue research
inspect the original source
search for contradictory evidence
answer
park
escalate
Evidence Display In Observability
Source visibility should connect to the MWMS AI Observability Metadata Standard.
Observability should record:
claim count
source count
inspected source count
failed source inspections
original sources found
derivative sources found
independent evidence chains
source trust rating
source freshness rating
source used in answer
evidence role
source conflict status
evidence sufficiency
verdict
confidence
This gives HeadOffice visibility into evidence quality across AI Employees.
Evidence Display In Evaluation Scorecards
Source visibility supports the MWMS AI Employee Evaluation Scorecard Standard.
Source display helps evaluate:
claim precision
source quality
source independence
freshness
factuality
answer relevancy
contradiction handling
confidence calibration
decision usefulness
traceability
safety
If source display is weak, evaluation quality is weaker.
Source Display And Confidence
Confidence must reflect evidence quality.
Confidence should be reduced when:
sources are low trust
source freshness is unknown
no original source was found
no independent source was found where required
sources conflict
source inspection failed
evidence is indirect
only commercial sources were used
only user-generated sources were used
apparently separate sources share one evidence base
source summaries are incomplete
final answer relies on assumptions
credible minority evidence remains unresolved
Rule
A high-confidence answer requires visible and sufficient evidence.
A high source count does not justify high confidence.
Source Display And Human Review
Human review should be triggered when:
source evidence is weak
sources conflict
freshness is unknown on a current topic
only low-trust sources are available
source inspection fails
source independence is unclear
the original source cannot be found
final answer affects compliance
final answer affects budget
final answer affects campaign launch
final answer is client-facing
evidence is insufficient for the decision
credible disagreement remains
the verdict is high impact
The review panel should show the sources and evidence conditions that caused concern.
Source Display And Kaizen
Weak source visibility should create Kaizen learning.
Kaizen notes may include:
claim not clearly displayed
source type classification missing
freshness not captured
trust rating missing
evidence role missing
source summary too vague
source independence not checked
derivative source presented as original
source conflict not shown
source used in answer not marked
evidence sufficiency unclear
verdict and confidence confused
source UI difficult to review
operator could not tell why the answer was trusted
These may become future improvements or regression tests.
Source Display Failure Conditions
Source visibility should be marked failed or weak if:
sources were used but not displayed
claim being evaluated is unclear
source URLs or record references are missing
source titles are missing
source summaries are missing
evidence roles are missing
freshness is missing where required
trust rating is missing where required
source independence is ignored
derivative sources are counted as independent confirmation
source conflict is hidden
unresolved disagreement is hidden
source inspection failed but the answer appears confident
final answer cites evidence that is not stored
source list is too vague to audit
operator cannot tell which source supported which claim
verdict and confidence are merged
interpretation is presented as source fact
search snippets are presented as inspected evidence
Minimum Starting Implementation
MWMS does not need a perfect source display UI immediately.
Minimum starting source display should include:
claim or question
source title
source URL or source record ID
source type
retrieved date
evidence role
evidence summary
freshness rating where relevant
trust rating where relevant
independence status where relevant
used in final answer status
source inspection status
limitations
conflict status
This is enough to move from hidden evidence to visible evidence.
Recommended Claim Display Object
{
“claim_id”: “”,
“original_claim”: “”,
“normalised_claim”: “”,
“claim_type”: “”,
“claim_origin”: “”,
“originating_source_id”: “”,
“claim_date”: “”,
“time_sensitivity”: “”,
“verdict”: “”,
“confidence”: “”,
“evidence_sufficiency”: “”,
“supporting_evidence_ids”: [],
“challenging_evidence_ids”: [],
“context_evidence_ids”: [],
“missing_evidence”: [],
“unresolved_issues”: []
}
Recommended Source Display Object
{
“source_id”: “”,
“source_title”: “”,
“source_url”: “”,
“source_type”: “”,
“source_owner”: “”,
“original_source_id”: “”,
“derived_from_source_id”: “”,
“publication_date”: “”,
“event_date”: “”,
“last_updated_date”: “”,
“retrieved_at”: “”,
“freshness_rating”: “”,
“trust_rating”: “”,
“relevance_rating”: “”,
“evidence_role”: “”,
“evidence_summary”: “”,
“key_claims”: [],
“claim_ids”: [],
“limitations”: [],
“independent_evidence”: “”,
“shared_evidence_base”: “”,
“used_in_final_output”: false,
“conflict_detected”: false,
“source_status”: “”,
“verification_status”: “”
}
Recommended Evidence Display Object
{
“evidence_id”: “”,
“claim_id”: “”,
“source_id”: “”,
“evidence_summary”: “”,
“evidence_location”: “”,
“evidence_role”: “”,
“evidence_strength”: “”,
“source_reliability”: “”,
“independent_evidence”: “”,
“contradiction_status”: “”,
“used_in_final_conclusion”: false
}
These are conceptual objects only.
Exact implementation may be adapted later.
Relationship To Research Planning And Query Rewriting
Research planning defines what sources and evidence should be found.
Source visibility shows what sources and evidence were actually found and used.
The research plan may say:
Find the official Google Ads policy source.
Source visibility should show:
whether that source was found
whether it was inspected
whether it was current
which claim it relates to
whether it supported the answer
what limitations remained
This closes the loop between planning and evidence.
Relationship To Deep Search Quality And Observability
The MWMS Deep Search Quality And Observability Framework governs:
claim definition
retrieval
evidence quality
source independence
contradiction handling
verdict quality
confidence calibration
workflow observability
This standard governs how those elements are displayed to human operators.
Relationship To Independent Model Review
The MWMS Independent Model Review And Rescue Routing Framework may use several models to review an output.
This standard requires the display to show:
which model used which sources
whether evidence was independently retrieved
whether models shared the same evidence base
where models agreed
where models disagreed
what minority evidence remained
Rule
Model consensus must not be displayed as evidence consensus without source-level support.
Relationship To Research Synthesis Documentation And Distribution
The MWMS Research Synthesis Documentation And Distribution Framework governs how research moves into a distributable report.
This standard ensures distributed reports preserve:
source visibility
claim-to-evidence mapping
checked findings
supporting and challenging evidence
limitations
unresolved disagreement
verdict
confidence
Relationship To Search Scrape Summarise Pipeline
This standard points toward a future MWMS Search Scrape Summarise Evidence Pipeline Standard.
That future standard may define the operational pipeline for:
searching
inspecting
summarising
storing
rating
linking
displaying evidence
This page defines the display and governance expectations.
Relationship To Guardrail And Preflight Checks
The AI Guardrail And Preflight Check Standard decides whether sources are required before work begins.
This page defines how those sources should be displayed once used.
Preflight decides:
Are sources required?
Research planning decides:
What sources should be found?
Deep Search decides:
How should sources be inspected and evaluated?
Source visibility decides:
What sources were used, what did they prove, and are they good enough?
Future Enhancements
Future enhancements may include:
MWMS Search Scrape Summarise Evidence Pipeline Standard
MWMS Deep Search Source Record Standard
MWMS Source Summary Record Schema
MWMS Claim And Evidence Record Schema
MWMS Source Independence Graph
MWMS Evidence Sufficiency Scorecard
MWMS Source Display UI Component Specification
MWMS HeadOffice Evidence Review Panel Specification
MWMS Client Facing Evidence Display Standard
MWMS Research Brain Source Library
These should be created only when implementation or system complexity justifies them.
Drift Protection
This standard prevents the following drift:
hiding sources from operators
treating links as enough evidence
showing citations without trust context
showing source lists without claim connections
using stale sources without warning
using commercial sources as proof
counting derivative reports as independent sources
ignoring source conflicts
hiding unresolved disagreement
hiding failed source inspection
giving high confidence with weak evidence
combining verdict and confidence
presenting interpretation as source fact
separating final answers from supporting evidence
making research impossible to audit
making client-facing outputs difficult to trust
losing source quality signals from AI work sessions
If sources matter, they must be visible.
If evidence is weak, the operator must be able to see why.
Drift Signals
Watch for:
“There are plenty of links.”
“Three articles say the same thing.”
“They all repeat the company announcement, but that is fine.”
“The citation proves the claim.”
“The source looks credible.”
“The AI is confident.”
“We do not need the original report.”
“The disagreement will confuse the reader.”
“Put all the sources at the end.”
“The verdict and confidence are basically the same.”
“The search snippet gives us enough.”
“We inspected similar sources.”
“The models all agreed.”
Rule
When these drift signals appear, return to claim visibility, evidence roles, original sources, source independence, conflict display, verdict discipline, and confidence calibration.
Architectural Intent
The architectural intent of this standard is to make MWMS evidence-first.
MWMS is not building AI Employees that simply answer.
MWMS is building AI Employees that show:
what was evaluated
what evidence was used
what each source proved
what each source did not prove
where disagreement remains
how confident MWMS should be
For MWMS to trust AI-generated intelligence, the system must display the evidence behind the answer in a way humans can inspect, challenge, review, and improve.
Source visibility turns AI research into operational intelligence.
It makes the system:
more trustworthy
more auditable
more useful
more challengeable
more correctable
more ready for future client-facing delivery
Final Standard
The MWMS final standard is:
No material source-backed answer, research report, fact check, client intelligence report, compliance review, affiliate evaluation, AI recommendation, or Deep Search output should be treated as decision-ready unless the operator can see the claim, sources, evidence roles, source quality, freshness, independence, limitations, conflicts, verdict, confidence, and evidence sufficiency behind it.
A valid MWMS evidence display should define:
claim
claim type
source title
source location
source type
source owner
publication or event date
retrieval date
freshness
trust
relevance
evidence role
evidence summary
claim relationship
original source
source independence
source status
verification status
limitations
conflicts
supporting evidence
challenging evidence
unresolved disagreement
verdict
confidence
evidence sufficiency
human review requirement
That is the MWMS Source Visibility And Evidence Display Standard.
MWMS System Change Log
Version: v1.1
Date: 2026-06-21
Author: HeadOffice
Change
Updated the MWMS Source Visibility And Evidence Display Standard from v1.0 to v1.1 using the AI Automations by Jack material covering:
browser fact-checking interfaces
claim verification
source-backed client research
multi-model research
source comparison
supporting and opposing evidence
confidence display
verdict display
source URL preservation
fact-checking copilots
research report evidence presentation
Expanded the standard beyond general source cards and source-quality indicators to include claim-level evidence display and source-relationship governance.
Added new standards covering:
• claim visibility
• claim types
• claim-to-evidence mapping
• evidence roles
• evidence summaries versus source descriptions
• source independence
• original-source preservation
• derivative-source identification
• original claims versus independent verification
• unresolved disagreement display
• controlled verdict display
• verdict versus explanation separation
• confidence versus truth separation
• source-list theatre prevention
• search-snippet limitations
• multi-model evidence display
Added recommended conceptual objects for:
• Claim Display
• Source Display
• Evidence Display
Expanded the Source Display Object to include:
• Original Source ID
• Derived From Source ID
• Event Date
• Evidence Role
• Claim IDs
• Independent Evidence
• Shared Evidence Base
• Verification Status
Added explicit doctrine that:
• a citation is not automatically supporting evidence
• a company statement proves that the company made the statement but does not independently prove the statement is true
• multiple URLs may represent one evidence chain
• model consensus is not evidence consensus
• verdict and confidence must remain separate
• unresolved disagreement must not be hidden
Change Impact Declaration
This update materially strengthens the source and evidence display standard without changing its primary authority.
HeadOffice remains the authority owner.
Research Brain remains responsible for source and evidence quality.
Data Brain remains responsible for structured source, claim, and evidence records.
Product Brain and UX Brain may later implement evidence display interfaces, but they do not determine source truth.
The update does not require every low-risk answer to display a complex evidence panel.
Display depth should match:
• task risk
• claim importance
• audience
• distribution type
• decision impact
The update does require material outputs to preserve enough evidence visibility for review.
Pages Created
• None
Pages Updated
• MWMS Source Visibility And Evidence Display Standard updated from v1.0 to v1.1
Pages Deprecated
• None
Standalone Pages Not Created
The following standalone pages were not created because their durable intelligence is governed within this updated standard:
• MWMS Claim Evidence Display Standard
• MWMS Fact Check Evidence UI Standard
• MWMS Source Independence Display Standard
• MWMS Multi Model Evidence Display Standard
• MWMS Source Conflict Display Standard
• MWMS Verdict And Confidence Display Standard
• MWMS Original Source Preservation Standard
• MWMS Claim To Evidence Mapping Standard
Registries Requiring Update
• MCR Page Registry
• HeadOffice Page Registry
• Research Brain Page Registry where this standard is operationally referenced
• MCR Copy Map where the framework copy and version are recorded
• MWMS Course Absorption Decision Registry
Canon Version Update Required
No immediate HeadOffice Canon or Research Brain Canon version change is required unless either Canon directly records source-display versions or contains source-visibility rules that conflict with v1.1.
The new claim, evidence-role, source-independence, verdict, confidence, and disagreement controls should be included during the next scheduled HeadOffice and Research Brain Canon alignment review.
Change Log Entry Required
Yes.
The v1.1 update must be recorded in:
• MWMS System Change Log
• MCR Page Registry change history where applicable
• HeadOffice Page Registry change history where applicable
• Research Brain Page Registry change history where applicable
• MWMS Course Absorption Decision Registry
Strategic Absorption Result
The AI Automations by Jack material concerning browser fact-checking, multi-model research, source verification, client intelligence research, verdict generation, confidence display, and evidence-backed reporting has been absorbed into the existing MWMS Source Visibility And Evidence Display Standard.
The absorption preserves the durable evidence-display intelligence while rejecting:
• citation theatre
• source-count theatre
• repeated reports being treated as independent evidence
• search snippets being presented as inspected evidence
• confidence being presented as truth
• model agreement being presented as factual verification
• hidden contradictory evidence
• unresolved disagreement being removed for presentation quality
• source lists disconnected from individual claims
The resulting v1.1 standard establishes that MWMS evidence display must be:
• claim-specific
• source-visible
• evidence-role aware
• original-source aware
• independence-aware
• freshness-aware
• conflict-visible
• disagreement-preserving
• verdict-controlled
• confidence-calibrated
• privacy-safe
• operator-reviewable
END OF FULL FILE OUTPUT