The New Newsroom · VERITAS Platform · Complete Documentation · June 2026
VERITAS
AI-Augmented Truth Infrastructure for Global Health Knowledge — SDG-guided newsroom to platform architecture, built for journalists.
01 · SDG Newsroom Playbook
The New Newsroom: SDG Intelligence
RAG architecture · Five formats · SDG gap mapping · Agent stack · Day in the life
02 · AI Journalist Playbook
VERITAS: AI-Augmented Journalist
Five modules · Investigation workflow · Story formats · Truth standards
03 · Platform UX
VERITAS: Platform Architecture
Investigation workspace · Social formats · Long-form article editor
Wellcome Trust Grant Application · Africa · Asia · South America
Infectious Disease · Mental Health · Climate & Health · Discovery Research
GraphRAG · Zero Hallucination · Five Output Formats · NewsGuard Compliant
01 · SDG Newsroom Playbook
The New Newsroom — SDG Intelligence
↑ Contents
SDG Intelligence Newsroom · AI Playbook v2.0

The New
Newsroom

How experienced journalists use RAG-grounded AI and UN SDG intelligence to produce verified, multi-format news — from academic research to global audiences — without sacrificing a single editorial standard.

Zero hallucination
RAG · GraphRAG · SDG
————
Reels · Live · Wire
Radio · Long-form
————
NewsGuard compliant
00
The Mission
Bridging peer-reviewed research and global audiences, guided by SDGs, grounded by RAG
Why this exists

The world's most important research — the peer-reviewed studies that contain real solutions to climate breakdown, food insecurity, and inequality — never reaches the people it could help. The gap between academic publication and public understanding is a structural failure. This newsroom exists to close it, using AI as the translation and distribution engine, and experienced journalists as the editorial intelligence that makes it trustworthy.

Architecture

Zero hallucination by design

An isolated RAG framework means the AI can only use what the researcher provided. No outside facts. No assumed context. Everything in the output traces back to the source paper.

SDG Intelligence

Research matched to global need

Every paper is scanned against the 17 UN Sustainable Development Goals and global coverage gaps. The AI identifies which findings are most undercovered and most urgent for world media.

Editorial Authority

Journalists decide everything that matters

What gets covered. What angle serves the public. Who to contact. Whether to publish. AI does not touch these decisions. It works for the editor — not the other way around.

The critical upgrade in v2.0

The original approach used flat RAG — retrieving isolated sentences from the paper. Version 2.0 uses GraphRAG, which builds a knowledge graph of the paper's entities, claims, and evidence chains. This means the traceability layer doesn't just cite a page number — it maps the full chain of evidence from conclusion back to raw data.

Original approach — gaps
Flat RAG retrieves isolated sentences
No flag for contested or preliminary findings
Two output formats (Reuters, Economist) — global north audience only
Active pitching without partner relevance scoring
SDG used as a filter — not as a live editorial signal
Version 2.0 — enhancements
GraphRAG traces full evidence chains
Confidence layer flags findings vs limitations
Five output registers including community and radio
Pitching agent scores by partner beat and geography
SDG gap analysis drives editorial commissioning
01
RAG Architecture
How the zero-hallucination framework works — and where it needs reinforcing
The Engine

Retrieval-Augmented Generation grounds AI output in a specific, curated knowledge source rather than the model's training data. For journalism, this is the only responsible architecture — because it means the AI cannot invent, assume, or embellish. It can only work with what you give it.

Flat RAG — original approach
Source
Academic paper (PDF / structured text)
Chunk + Embed
Paper split into text chunks, vectorised
Retrieve
Query finds relevant chunks
Generate
Output grounded in retrieved chunks
GraphRAG — v2.0 upgrade
Source
Academic paper with full citation graph
Knowledge Graph
Entities, claims & evidence chains extracted
Confidence Layer
Flags: findings / methods / limitations
Traced Output
Every sentence mapped to evidence chain
The failure mode RAG prevents

Standard LLMs hallucinate — they produce fluent, confident text that is factually wrong. A 2025 BBC study found over half of AI answers to news queries had significant issues including factual errors and fabricated quotations. RAG with a strictly isolated knowledge source removes this failure mode entirely, because the model has nowhere else to look.

Why GraphRAG matters for academic papers

Academic papers are not flat documents. They contain linked claims: a conclusion in section 5 depends on a method in section 2, which cites data in an appendix. Flat RAG retrieves the conclusion without the chain. GraphRAG retrieves the whole chain — so the traceability view shows the editor not just where the claim appears but what it depends on. This is the difference between citation and verification.

Confidence layer — new

Three-tier claim classification

Every claim in the output is tagged as drawn from: Core findings (highest confidence), Methodology (conditional), or Limitations (requires explicit qualification in the news piece). The editor sees these tags in the side-by-side view before approving.

Conflict detection — new

Contested findings flag

When the paper's own data presents tension — a finding that contradicts prior literature it cites, or a limitation that qualifies a headline result — the RAG system flags this to the editor. It does not resolve the dispute. It routes it to human judgment.

02
SDG Intelligence
Using the UN's 17 goals as a live editorial gap map, not just a topic filter
The Editorial Compass

The UN SDG Media Compact now has almost 400 member organisations across 160 countries reaching a combined audience of two billion people — all actively seeking content that connects news to the Goals. This newsroom doesn't just reference the SDGs. It uses them as a real-time signal for where verified, factual journalism is most needed and least available.

SDG gap analysis — the editorial upgrade

Rather than scanning academic papers for SDG connections after the fact, the SDG intelligence layer runs continuously. It tracks which of the 17 Goals has the least verified news coverage in the past 30–90 days, broken down by region. This becomes the primary signal the editor uses when commissioning academic sources — turning research selection from reactive to strategic.

The 17 Goals — active coverage tracking

Coverage gap

Which Goals get underreported

SDGs 2 (Zero Hunger), 6 (Clean Water), 14 (Life Below Water) and 16 (Peace & Justice) consistently receive a fraction of the media coverage that SDG 13 (Climate) attracts. The gap signal prioritises these.

Regional targeting

Coverage by geography

A story on SDG 3 (Good Health) is well-covered in Western Europe but severely underreported across Sub-Saharan Africa. The SDG layer identifies the regional gap, not just the topical one.

Commissioning signal

Drives which academics to approach

The editor doesn't wait for papers to arrive. The SDG gap map shows which research topics are most needed right now — then the editor finds the right academic to match.

03
Agent Stack
The four AI agents and what each one does — and doesn't do

Each agent has a strictly bounded job. None of them make editorial decisions. All of them report to the editor. The isolation is intentional — it prevents scope creep where an agent starts making judgments it has no authority to make.

R

Research Agent — SDG gap scanner + signal detector

Scans the academic paper against the live SDG coverage gap map and historical global news patterns. Identifies the 3–5 most impactful narrative angles. Flags which regions and media markets need this research most. Outputs a ranked editorial brief — not a story, a brief.

GraphRAG · SDG gap index · NewsWhip signal data
L

Language Condensation Agent — jargon removal + style translation

Strips academic density and rewrites for five audience registers: Reuters wire style, Economist analysis, BBC World Service radio, plain-language community brief, and social-native Reel script. The RAG constraint means every claim in every version traces to the source paper. No outside colour added.

Isolated RAG · Five style guides · Confidence layer tags visible Limitations-section claims flagged in all outputs
T

Traceability Agent — sentence-level citation mapping

Maps every sentence in every output format back to the exact page, paragraph, and claim tier in the original paper. Produces the side-by-side view the academic sees before final approval. Also generates a NewsGuard compliance report showing the evidence chain for every published claim.

GraphRAG citation graph · Three-tier confidence display · NewsGuard export
D

Distribution Agent — relevance-scored pitching

Monitors breaking global news. When a relevant event occurs, it matches pre-translated packages to partner news desks by SDG beat, geographic coverage area, and language preference — then pitches only to editors for whom this is directly relevant. Not a broadcast. A targeted, contextual pitch.

Partner beat database · Breaking news feed · Multi-language packages No pitch sent without relevance score above threshold
What none of the agents do

None of the agents decide what gets covered. None contact sources directly. None approve final copy. None bypass the academic's review of the traceability map. None publish without the editor's sign-off. Every agent output is a draft routed to human judgment.

04
Five Formats
One verified research story packaged for five distinct audiences simultaneously

The original approach produced three formats. Version 2.0 produces five — adding a community plain-language brief and a BBC-style radio script, which are the formats that reach the communities most directly affected by SDG issues. All five derive from the same RAG-grounded, traceability-mapped source material.

Format Audience Style register Length Distribution channel
Digital wire copyFor editors at global outlets Global news desks Reuters style 400–600 words Active pitch via distribution agent
Analysis featureFor quality press and magazines Informed general readers Economist style 900–1400 words Long-form partners, newsletters
Radio scriptNew — regional broadcasters Radio audiences, Global South BBC World Service 3–4 min read Regional broadcast partners
Community briefNew — NGOs, schools, local orgs Affected communities directly Plain English Sub-800 words NGO partner network, SDG Compact
Social packageReel script + caption + thread Social media audiences Platform-native 60–90s script + captions Own channels + partner reposts
The one-story rule

All five formats are generated from a single RAG-grounded condensation of the source paper. The language condensation agent applies the five style guides in parallel — it does not rewrite the story five times independently. This means the factual core is identical across all formats, and the traceability map covers all five simultaneously.

Why radio and community brief matter

Reaching the communities SDGs are about

SDG stories about hunger, water access, and poverty need to reach the communities experiencing those issues. Wire copy reaches editors in capital cities. Radio reaches rural communities without reliable internet. Community briefs reach local NGO workers and teachers. These are the audiences with the highest stake in the research.

Academic approval covers all five

One review, five sign-offs

The academic sees the traceability map for all five formats simultaneously. They approve the factual core once — the style differences between Reuters copy and a community brief are not their concern. Their sign-off is that every claim in every format is accurate to their research.

05
Distribution
Active, relevance-scored pitching — not broadcasting into the void

The original active pitching concept is strong but needs precision. A news desk that receives an unsolicited research package when a climate crisis hits will ignore it unless it's immediately, specifically relevant to their beat, geography, and audience. The distribution agent's job is to know the difference.

1

Breaking news signal monitored continuously

The distribution agent tracks global news feeds for events that match the SDG tags and regional markers attached to every pre-translated package in the library.

NewsWhip · AP wire monitor · UN news feed
2

Partner beat database — who covers what, where

Every media partner in the network has a profile: SDG beats they cover, geographic regions, language, audience type (general, specialist, community), and their preferred format (wire, radio script, briefing). The pitching agent matches against all four dimensions before sending anything.

Partner CRM · SDG beat tags · Format preference matrix
3

Relevance scoring — threshold before any pitch is sent

A package is only pitched to a partner if its relevance score — calculated from SDG match, geographic overlap, recency, and breaking news context — exceeds a set threshold. Below threshold: the match is flagged for the editor to review manually before deciding whether to pitch.

Relevance scoring model · Editor review queue for borderline matches No automated pitch below relevance threshold
4

Editor approves all pitches above threshold

Even high-scoring matches are reviewed by the editor before the pitch goes out. The agent prepares the pitch; the editor sends it. This keeps the relationship between this newsroom and its partners in human hands.

Multi-format distribution by channel type

Wire copy goes to news desk editors. Radio scripts go to broadcast producers. Community briefs go to NGO programme officers and school coordinators. Social packages go to digital editors and social media managers. The same package, delivered in the right format, to the right person, at the right moment — is the difference between pickup and deletion.

06
The Editor
What no agent touches — and why that's the whole point

The editor is not a gatekeeper at the end of an AI pipeline. They are the intelligence that shapes the pipeline from the start. The SDG gap map informs their commissioning. The research brief informs their story selection. But the judgment — what matters, why it matters now, who needs to hear it — is entirely theirs.

Commissioning

SDG gap map drives what gets covered

The editor uses the live SDG coverage gap analysis to identify which Goals, in which regions, are most underserved by factual journalism right now. This shapes which academic researchers they approach — making editorial selection strategic rather than reactive.

Sourcing

Finding and briefing the researcher

The editor identifies researchers who have already published peer-reviewed work on the selected SDG gap. They interview the academic not just to understand the paper, but to extract the human story — the real-world application the data points to — before anything goes to the AI.

Collaborative bridge

Translating research intent before AI touches it

The editor's conversation with the academic determines the core message. What did this research actually find? What should a policymaker, a teacher, a farmer do differently because of this? This framing is set by humans before any agent is given the paper.

Final authority

Every publish decision, every pitch, every correction

The editor approves all five format outputs. They approve every distribution pitch. They issue all corrections — publicly, on every platform, at the same prominence as the original. The accountability chain ends with them, always.

Why the editor-as-bridge role is the most important job in this newsroom

The RAG system can only be as good as the understanding of the research that was fed into it. If the editor's briefing conversation with the academic is shallow, the research agent's brief will be shallow, and the language condensation agent will faithfully produce shallow wire copy. The depth of the journalism flows from this conversation — not from the AI. This is the role the technology cannot touch.

07
Truth Rules
Non-negotiable — across every format, every platform, every deadline

The RAG isolation and GraphRAG traceability are architectural truth protections. These rules are human truth protections. They apply regardless of how good the technology is, because technology does not bear reputational accountability. You do.

The Eight Rules

01 RAG output is not a finished fact. It is a draft grounded in a source. Every claim still requires the editor's judgment that the source itself is credible — a peer-reviewed paper with contested methodology is not the same as one with replicated results.
02 Limitations-section claims must be qualified in every format. When the traceability agent flags a claim as drawn from the paper's limitations section, that qualification must appear in the published text. "The researchers note this finding is preliminary and based on a limited sample" is not optional hedging — it is accurate reporting.
03 The academic's approval is not a publication approval. The researcher confirms the traceability map is accurate. The editor decides whether the story serves the public interest and is ready to publish. These are different decisions made by different people.
04 Format does not change the standard. A 60-second Reel carries identical editorial responsibility to a 1,200-word wire story. The RAG traceability map covers all five formats. So does your accountability.
05 Corrections are public, immediate, and prominent. If any format contains an error — whether in the wire copy or a social caption — the correction appears on every platform at the same prominence as the original. A buried correction is not a correction.
06 Disclose AI use in all published work. Every format should carry a brief disclosure: that AI was used to condense and reformat the research, that the RAG system was grounded exclusively in the source paper, and that the academic reviewed the traceability map. This transparency is what makes the NewsGuard compliance claim credible.
07 Breaking news relevance does not lower verification standards. When the distribution agent flags a story as urgent because a climate crisis just broke, the pitch may go out faster — but the five formats were already verified before the crisis hit. Nothing gets expedited through verification because of a news hook.
08 The editor is accountable for everything published under this newsroom's name. The agents, the RAG architecture, the GraphRAG knowledge graph — none of these share editorial accountability. The editor does. Every publication decision is theirs.
08
Day in the Life
From academic paper to five published formats — the full workflow

This timeline shows a single story moving through the complete system — from the editor's commissioning decision through to live distribution across five formats. It assumes the academic paper has already been delivered. The whole cycle, with an experienced editor and a functional agent stack, runs in roughly one working day.

08:00

SDG gap review — editorial commissioning Editor

Editor reviews overnight SDG coverage gap analysis. Identifies which Goals and regions are underserved. Selects the academic paper for today's cycle based on this signal.

09:00

Academic briefing call Editor + Researcher

Editor interviews the researcher to extract the core message — the human implication of the data. This conversation shapes what the research agent prioritises. Cannot be skipped or delegated.

10:00

Research agent runs — SDG angle brief Research Agent

GraphRAG builds the paper's knowledge graph. Research agent scans against SDG gap map and news history. Outputs ranked editorial brief with 3–5 narrative angles and target markets.

10:30

Editor selects angle and target markets Editor

Reviews the research brief. Selects the primary angle. Identifies which of the five formats to prioritise. Approves the brief before condensation begins.

11:00

Language condensation — all five formats in parallel RAG

Isolated RAG condenses the paper into five simultaneous drafts. Confidence layer tags every claim. Limitations-section findings flagged across all formats automatically.

11:45

Traceability map generated Traceability Agent

Every sentence in every format mapped to exact page, paragraph and claim tier in the source paper. Side-by-side view prepared for academic review. NewsGuard compliance report generated.

12:30

Academic reviews traceability map Researcher

Researcher reviews side-by-side view. Confirms every claim accurately represents their research. Raises any concerns about qualification of preliminary findings. Approves the factual core.

13:30

Editor final review and editorial additions Editor

Editor reviews all five formats. Adds context the RAG cannot provide — current events relevance, additional sourcing, editorial voice. Makes the public interest case. Final copy is edited, not just approved.

14:30

Publication — wire + social package live Publish

Wire copy and social package published first. Reel filmed and posted. Distribution agent begins matching against partner beat database.

15:00

Editor approves distribution pitches Editor

Distribution agent presents scored pitch list. Editor reviews, approves or adjusts each pitch. Pitches go to relevant news desks, radio producers, NGO partners.

16:00

Long-form analysis + community brief published Publish

Economist-style analysis and community plain-language brief go live. Radio script delivered to broadcast partners. All five formats now in distribution.

17:30

Monitoring, corrections, next commissioning signal Editor

Monitor partner pickup and audience response. Address any corrections immediately across all formats. Review SDG gap signal for tomorrow's commissioning decision.

What this workflow achieves that a traditional newsroom cannot

A traditional team of three journalists cannot produce five publication-ready formats from a complex academic paper in one working day while maintaining full source traceability and academic sign-off. The AI agents compress the production work. The editor focuses entirely on judgment, editorial depth, and relationships. That combination — machine speed, human standards — is what makes this newsroom different.

02 · AI Journalist Playbook
VERITAS — The AI-Augmented Journalist
↑ Contents
AI-Augmented Journalist Playbook · VERITAS Platform

VERITAS

One platform. Every verification function. Human judgment at the centre of every decision.
00
What is VERITAS
A unified truth infrastructure — not another standalone tool
The brief

VERITAS combines everything Bellingcat, ProPublica, Wikipedia, Full Fact and PolitiFact do — in a single workspace, with AI handling the volume work and journalists retaining every decision that matters. It exists because the verification crisis is not a shortage of tools. It is a shortage of integration.

Verify

Claims and media, at speed

GraphRAG searches vetted primary sources simultaneously. Synthetic media detection runs on every uploaded asset. Conflicting evidence is surfaced, not suppressed.

Investigate

Documents and OSINT, unified

OSINT enrichment, document intelligence, and knowledge graph building in one session. What takes a Bellingcat-trained investigator half a day takes minutes.

Research

Peer-reviewed evidence, accessible

The research bridge surfaces relevant academic literature inside the investigation workspace. Plain-language synthesis, confidence-tiered findings, full traceability.

The core difference

The competitive advantage is not any single capability — each function has existing counterparts. It is the workflow. A journalist never leaves the platform to complete an investigation. Every module shares the same evidence base, the same source graph, and the same traceability layer.

The fragmented world today
Bellingcat toolkit for OSINT — requires specialist training
ProPublica methods for documents — not replicable by most newsrooms
ClaimBuster for claim detection — standalone, no integration
InVid/WeVerify for video — separate plugin, separate workflow
Separate academic search for research context
No shared knowledge base between newsrooms
VERITAS — one session
Evidence workspace: OSINT + deepfake + metadata in one dossier
Document intelligence: any volume, any newsroom, no specialist required
Claim engine: GraphRAG across all primary sources simultaneously
Multimodal detection: integrated, confidence-scored, methodology visible
Research bridge: academic literature inside the investigation
Collaborative layer: shared, source-locked, attributed knowledge base
01
The AI-Augmented Journalist
More capability. Same authority. Zero compromise on standards.
The design philosophy

Augmentation means the journalist gains speed and capability without losing authority. AI handles the mechanical: scanning, transcription, cross-referencing, pattern detection. The journalist handles everything that requires judgment, context, source relationships, and ethical reasoning. The platform is designed so these distinctions are visible and structural — not aspirational.

TaskWhat AI doesWhat the journalist does
Evidence gatheringOSINT, metadata, imageryRuns all OSINT enrichment, reverse searches, metadata extraction, geolocation simultaneously. Returns confidence-scored dossier.Decides what the evidence means. Determines whether sources are credible. Judges whether to pursue the story.
Document analysisLarge document setsTranscribes, classifies, builds knowledge graph, surfaces patterns and anomalies, cross-references across documents.Directs what to look for. Interprets what the patterns mean. Decides what constitutes a finding worth publishing.
Claim verificationStatements and assertionsSearches vetted primary sources simultaneously. Surfaces supporting and conflicting evidence with full methodology.Reads the evidence landscape. Decides how to qualify the claim. Makes the editorial call on what the truth is.
Synthetic mediaImages, video, audioRuns multimodal detection, outputs confidence score and forensic indicators — not a verdict.Reviews indicators. Applies contextual knowledge. Decides whether to use the material or flag it.
Research synthesisAcademic literatureSurfaces relevant papers, classifies findings by confidence tier, produces plain-language synthesis with traceability.Evaluates methodology. Judges research quality. Decides which findings are appropriate to include in journalism.
PublicationAll formatsGenerates draft structure, format variants, caption options — all traceable to source material.Writes the story. Makes every editorial decision. Approves every published sentence. Owns full accountability.

The line that never moves

01AI surfaces evidence. Journalists interpret it. No VERITAS output is presented as a conclusion. Every finding includes its methodology and confidence score. The journalist decides what it means.
02AI detects. Journalists verify. A synthetic media flag is a prompt to investigate further, not a verdict. A claim match is evidence to weigh, not proof.
03AI drafts. Journalists write. Platform-generated structure and format drafts are scaffolds. The journalist's voice, judgment, and editorial decisions produce the published story.
04AI works for the journalist. Not the other way around. The platform is designed around the journalist's workflow, not the AI's output. If the AI slows down the work, the interface is wrong.
02
Five Modules
Each module addresses one siloed function — together they form a complete investigation workspace

Click any module to expand. Every module shares the same underlying evidence graph, so findings in one module are immediately accessible in all others.

01Evidence Workspace+

Drag in any asset — document, video, image, audio, URL. All enrichment runs simultaneously. The journalist receives a single dossier, not a list of tool outputs.

OSINT enrichment

Domain lookup, entity extraction, network mapping, geolocation cross-reference — all integrated, all sourced.

Reverse search

Image and video reverse search across multiple engines. Earliest verifiable appearance surfaced automatically.

Synthetic media detection

Multimodal AI forensics. Confidence score plus forensic indicators. Methodology always visible.

Metadata extraction

Full EXIF, creation date, device fingerprint, edit history — presented in plain language.

Better than Bellingcat because

Bellingcat's toolkit is a curated list of separate tools requiring specialist knowledge to operate. VERITAS integrates them, shares results across one evidence graph, and is designed for any working journalist — not only those with OSINT training.

02Claim Verification Engine+

Submit any claim — typed, pasted, or spoken. GraphRAG searches across the verified knowledge base: official databases, peer-reviewed research, government records, archived journalism, legal filings.

GraphRAG search

Searches across relational evidence graph, not flat text. Entity links and evidence chains retrieved, not just keyword matches.

Conflict surfacing

Where sources disagree, both are shown with equal prominence. No algorithmic verdict. The journalist sees the dispute.

Source hierarchy

Primary sources flagged above secondary. Peer-reviewed research flagged above news. Chain of evidence visible.

Confidence tiers

Core findings, methodology-dependent, and contested results shown in distinct tiers. No false certainty.

Better than Full Fact and PolitiFact because

VERITAS does not produce a single verdict with a rating label. It shows the journalist the evidence landscape — supporting and conflicting — so they can make their own call. Verdict-first systems embed editorial bias into the architecture. Evidence-first systems embed editorial authority into the journalist.

03Document Intelligence+

Upload any volume of documents — leaked files, FOIA releases, court records, financial filings. The platform transcribes, classifies, and builds a searchable knowledge graph. Available to any journalist, not just those with ProPublica-scale resources.

Bulk transcription

Audio, video, handwritten, scanned — all converted to searchable text. Accuracy flags on uncertain passages.

Knowledge graph

Entities, relationships, timelines, and anomalies extracted automatically. Cross-document connections surface without manual review.

Pattern detection

Statistical anomalies, repeated entities, unusual frequencies — flagged for journalist review.

Secure handling

End-to-end encryption. Self-hosted option for sensitive investigations. Source protection architecture built in.

Better than ProPublica's methods because

ProPublica's document tools are purpose-built for their own investigations and require significant technical implementation. VERITAS makes the same capability available to a reporter at a regional paper in Lagos or a freelance journalist in Bogotá — with no technical background required.

04Research Bridge+

Every investigation gets a live connection to peer-reviewed academic literature. The research bridge surfaces relevant papers, classifies findings by confidence tier, and produces plain-language synthesis — inside the same session the investigation is already in.

Relevance surfacing

Papers matched by entity, topic, and geographic context — not just keywords. Preprint flagged separately from peer-reviewed.

Confidence classification

Core findings vs methodology-dependent vs limitations-section. Journalist sees the tier before citing.

Plain-language synthesis

Jargon stripped. Key findings in journalistic English. Full citation visible at one click.

Traceability

Every synthesised sentence maps to its source paragraph. Academic can verify before citation.

05Collaborative Verification Layer+

A shared, source-locked knowledge base built by journalists, for journalists. Every contribution must cite a primary source. Every edit is attributed. Disputes are flagged, not resolved by majority vote.

Source-locked editing

No claim is added without a cited primary source. The architecture enforces this — it is not a policy.

Attribution

Every contribution is attributed to a verified journalist or researcher. No anonymous editing.

Dispute flagging

Where sources conflict, both versions remain visible with their citations. The dispute is surfaced, not arbitrated.

Living record

The knowledge base is citable, versioned, and permanently archived. A truth record that improves with use.

Better than Wikipedia because

Wikipedia can be edited by anyone, which creates reliability problems at scale. VERITAS's collaborative layer requires source citation by architecture, attributes every contribution to a verified professional, and flags disputes rather than resolving them by majority vote. Trust is built into the structure, not delegated to community governance.

03
Investigation Workflow
From incoming tip to published story — the VERITAS session

A VERITAS session follows the journalist's natural investigation arc. Each step builds on the last. Evidence gathered in Module 01 is automatically available in the claim engine. Documents processed in Module 03 feed into the collaborative layer. Nothing is siloed. Nothing is lost between tools.

TIP

Incoming tip or editorial signal

The journalist receives a tip, observes a social media claim, or the platform's monitoring surfaces an anomaly. This is the human editorial call — is this worth investigating? VERITAS does not make this decision.

01

Evidence workspace — first pass

Journalist drags the relevant assets into the evidence workspace. OSINT enrichment, metadata extraction, reverse searches, and synthetic media detection run simultaneously. Results arrive as a dossier with confidence scores and full methodology.

Evidence Workspace

Journalist reviews dossier — decides whether to proceed

The journalist reads the evidence. They decide what it means, whether the sources are credible, and whether the story is worth pursuing. AI found the needle. The journalist decides if it is the right needle.

02

Claim verification — key assertions checked

The journalist submits the key claims to the verification engine. GraphRAG returns the evidence landscape — supporting evidence, conflicting evidence, source confidence tiers. The journalist sees the full picture.

Claim EngineConflicts always surfaced
03

Document intelligence — if documents are involved

If the story involves document sets, they are uploaded and processed. Knowledge graph builds. Patterns and anomalies surface. Cross-document connections flagged for journalist review.

Document Intelligence
04

Research bridge — academic context

Relevant peer-reviewed research surfaces automatically from the investigation context. Journalist reviews confidence tiers, checks methodology, decides which findings are appropriate to cite.

Research Bridge

Source calls and original reporting

Everything AI cannot do. The journalist contacts sources, conducts interviews, gathers original quotes and responses. This step cannot be compressed or delegated. It is the journalism.

Format generation — all outputs from one evidence base

The platform generates draft structures for each required format — long article, social package, live brief — all traceable to the same verified evidence. The journalist writes, edits, and approves every published sentence.

Formats Module
PUB

Publication — journalist accountable for everything

The story publishes. Every claim in every format traces to verified evidence. The journalist's name is on it. The accountability is theirs. Entirely.

04
Story Formats
One investigation. Multiple formats. Same evidence base. Same standards.

VERITAS generates format scaffolds for three output types simultaneously from the same verified evidence session. The journalist selects, writes, and approves each. The traceability layer covers all three.

Short-form · Reels

The verified hook

AI identifies the single most evidentially surprising finding. Suggests a 3-beat 60-second script structure. Journalist films, rewrites, and approves. Every claim pre-verified before camera rolls.

Live · Multi-channel

The verified briefing

AI prepares pre-show briefing from the investigation's evidence base. Anticipated questions answered from verified findings. Live transcript auto-processed for clips. Only verified claims stated on air.

Long-form · Article

The complete record

AI generates structured skeleton from verified evidence graph. Journalist writes the story — context, quotes, analysis, meaning. Every AI sentence traceable to primary source. Full citation layer visible.


Short-form — the three-beat reel
0–3s

The hook — one arresting verified fact

AI surfaces the most evidentially striking finding from the investigation. Journalist selects and reframes. This sentence must be independently verified before filming. No exceptions.

3–50s

The evidence — two or three verified facts in sequence

Each fact earns the next. All drawn from the investigation's verified evidence base. Confidence tiers visible to journalist throughout. Limitations-section findings qualified explicitly.

end

The implication — what does this mean

The journalist's editorial voice. What should the audience do with this? Points toward the long-form investigation for depth. This section is human-written, always.


Long-form — the article architecture
SectionAI roleJournalist role
Lede and introSuggests opening fact from evidence graph, three variant approachesWrites the lede. Every word is theirs. The suggestion is a prompt, not a draft.
Evidence sectionsGenerates structured scaffold from verified findings, confidence-tiered, fully traceableWrites every sentence. Adds context, meaning, and narrative. Integrates original quotes.
Research contextSurfaces relevant peer-reviewed synthesis from research bridgeEvaluates and selects. Qualifies preliminary findings. Decides what belongs in the story.
Response and balanceFlags where named parties should be contacted for response based on evidence foundContacts sources. Obtains responses. Decides how to include them.
Headline variantsGenerates 8–10 headline options from the story's core findingSelects and edits. The final headline is always human-written.
The short-form trap — never shortcut verification

A 45-second Reel reaches the same audience as a 2,000-word article and carries identical editorial accountability. The evidence base is the same. The verification standard is the same. The speed of production does not compress the standard. It never has. It never will.

05
Truth Standards
Non-negotiable across every module, every format, every deadline

These standards apply whether you are using the evidence workspace or the social package generator, whether you have two hours or twenty minutes. The platform is designed to make following them faster — never to make bypassing them easier.

The eight commitments

01AI output is evidence, not conclusion. Every VERITAS finding includes its methodology and confidence score. The journalist decides what it means. The platform never presents a verdict.
02Conflicting evidence is always shown. When sources disagree, both appear with equal prominence. No algorithmic verdict. The journalist sees the dispute in full.
03Limitations-section findings are qualified. When a finding comes from a paper's limitations section, that qualification appears in the published story. This is not optional hedging. It is accurate reporting.
04Format does not change the standard. Reel, live broadcast, wire copy, long article — identical evidence standard, identical editorial accountability. Speed of production is irrelevant.
05Synthetic media flags require investigation, not publication. A deepfake detection flag is a prompt to investigate further. It is not proof of manipulation. It is not grounds to ignore the content either.
06Corrections are public, prominent, and immediate. When something is wrong, the correction goes on every platform, at the same prominence as the original. A buried correction is not a correction.
07AI use is disclosed. Every published story using VERITAS carries a brief disclosure of AI assistance and a link to the methodology. This transparency is the platform's credibility.
08The journalist is accountable for everything under their byline. VERITAS, GraphRAG, automated enrichment — none of these share editorial accountability. You do.
06
Day in the Life
A complete investigation cycle — tip to publication across three formats

This is what a full investigation day looks like with VERITAS — from incoming signal through verification to publication as a Reel, a live broadcast, and a long-form article. For a journalist or team of two to three.

08:00

Signal review and editorial decision Human

Review overnight monitoring alerts. VERITAS surfaces anomalies from connected feeds. Journalist decides what is worth investigating. AI found the signal. The editorial call is entirely yours.

08:45

Evidence workspace — first pass AI-assisted

Drop the relevant assets into the evidence workspace. OSINT enrichment, metadata, reverse search, synthetic detection all run. Dossier ready in minutes. Journalist reviews methodology and confidence scores.

09:30

Claim verification and document processing AI-assisted

Key claims submitted to the verification engine. Any documents uploaded to document intelligence. Knowledge graph builds. Conflicting evidence surfaced. Journalist reviews the full evidence landscape.

10:30

Original reporting — source calls Human only

Everything AI cannot do. The journalist calls sources, conducts interviews, seeks responses from named parties. Gathers original quotes. This step cannot be compressed. It is the journalism.

12:30

Reel — first publication Publish

Research bridge surfaces relevant academic context. AI identifies the evidential hook. Journalist films and approves verified 60-second Reel. Caption and platform package generated. Goes live.

14:00

Long-form article — writing phase AI scaffold + human writing

AI generates structure from verified evidence graph. Journalist writes every sentence — context, quotes, analysis, meaning. Every AI-generated scaffold sentence traced to primary source before inclusion.

15:30

Live broadcast — multi-channel Live

Pre-show briefing prepared from investigation's evidence base. Live across YouTube, Instagram, LinkedIn. Only verified claims stated on air. Real-time transcript processed for post-show clips.

17:00

Final article review and publication Human

Final verification pass. Every claim sourced. Every quote confirmed. AI use disclosed. Long-form article published with full source traceability. Collaborative layer updated with verified findings.

17:30

Derivative formats and monitoring AI-assisted

Post-show clips identified from live transcript. Newsletter summary generated. Social thread drafted. Journalist approves each before posting. Platform monitors for corrections signals overnight.

What VERITAS changes

The research phase is faster. The derivative format generation is faster. The briefing preparation is faster. The original reporting — source calls, verification judgment, and writing with meaning — takes exactly as long as it always has. VERITAS compresses the hours around the journalism. Not the journalism itself.

VERITAS · AI-Augmented Journalist PlaybookAI augments. Journalists decide.
03 · Platform Architecture & UX
VERITAS — Platform Architecture & UX
↑ Contents
GraphRAG · Active
Zero Hallucination
Evidence Workspace
Module 01 · OSINT + Detection
Current investigation
Government contract leak
14 assets · Started 09:14
In progress
Assets processed
leaked_contract_v3.pdf
Document · 47 pages
Processed
press_conference_clip.mp4
Video · 3m 22s
Synthetic: Clear
official_statement.jpg
Image · Metadata extracted
Review: Timestamp anomaly
company_filing_2024.pdf
Document · 12 pages
Cross-referenced
Claims verified: 6/9
Contract value: £4.2M
Core finding · 3 sources
Verified
Procurement bypassed
Conflict detected
Dispute: 2 sources
Drop evidence here
All enrichment runs simultaneously. Results arrive as a unified dossier.
PDF
Video
Image
Audio
URL
Document
Image · Metadata
⚠ Review
official_statement.jpg — timestamp anomaly
EXIF creation date: 14 March 2024. Document claims date: 22 March 2024. 8-day discrepancy detected. Earliest verifiable appearance: none found in archives prior to leak. Device fingerprint: iOS 17.2, iPhone 14.
Timestamp mismatch Confidence: 58%
Document · Cross-reference
✓ Verified
Contract value £4.2M confirmed
Value appears on pages 3, 18, and 41 of leaked document. Cross-referenced against Companies House filing ref CH-2024-4892 (£4.2M contract, same counterparty). Government procurement register: no matching record found.
3 sources Confidence: 94%
Claim Verification Engine
⚠ Conflicting evidence — journalist judgment required
Source A — Supports claim

Leaked contract contains no procurement authorisation signature on pages 12–14, where standard procedure requires sign-off. Companies House records show contract awarded 6 days before formal procurement window opened.

Source B — Disputes claim

Government spokesperson statement (archived 22 March 2024): "All procurement procedures were followed. Ministerial approval was obtained under emergency waiver provision S.14(b)." S.14(b) waiver records are not publicly accessible.

Academic Research Bridge
Core finding
Research on emergency procurement waivers in UK public contracts (2019–2024) found that S.14(b) waivers were invoked in 94% of cases reviewed without subsequent public disclosure, contradicting stated transparency commitments.
King's College London · Public Procurement Review · Vol 14 · 2024 · p.341
Every sentence mapped to source
"Contract value confirmed at £4.2 million"
leaked_contract_v3.pdf
Pages 3, 18, 41 · Companies House CH-2024-4892 Core
"No matching record in government procurement register"
procurement.gov.uk · Archive 23 Mar 2024
Verified via Wayback Machine Core
"Timestamp discrepancy of 8 days on official statement image"
official_statement.jpg · EXIF data
Requires further investigation Review
"Procurement bypassed — disputed by government"
Conflicting sources
Both sides must appear in published story Conflict
"S.14(b) waivers rarely disclosed publicly (94% of cases)"
KCL · Public Procurement Review · 2024
p.341 · Peer reviewed Core
Article structure
Headline options
Human writes · AI suggests
Lede
Human writes
Evidence: contract value
Verified · 3 sources
Evidence: register gap
Verified · archived
Conflict: waiver dispute
Both sides required
Research context
Research bridge
Response section
Awaiting: 3 parties
Conclusion
Human writes
Verification status
✓ 6 claims verified
⚠ 1 disputed claim
⏳ 2 awaiting response
Investigation · Public Spending
The £4.2 Million Contract That Isn't in the Register
Documents verified by VERITAS show a government contract awarded outside standard procurement procedures — and a disputed claim about the emergency powers used to justify it.
1 June 2026 AI-assisted · VERITAS · Full methodology
A government contract worth £4.2 million was awarded to [company name] without any corresponding entry in the public procurement register, according to documents independently verified by this newsroom across three separate sources.

The contract appears in a leaked internal document — verified for authenticity through metadata analysis, EXIF cross-checking, and device fingerprint — and matches a Companies House filing from March 2024. A search of the public procurement register, archived at three separate points between March and May 2024, contains no matching record.

"No matching entry exists in the public procurement register at the date of contract award."
Verified · procurement.gov.uk · Archived 23 March 2024 · Confidence: 94%
The government has disputed this characterisation, saying all procurement procedures were followed. A spokesperson cited the use of an emergency waiver under Section 14(b) of the Procurement Act, a provision that allows contracts to be awarded without the standard competitive process in defined emergency circumstances.

Records of S.14(b) waivers are not publicly accessible. This newsroom has been unable to independently verify or refute the government's account. Both accounts are presented here. Readers should weigh them accordingly.

Academic research on emergency procurement waivers in UK public contracts between 2019 and 2024 found that S.14(b) waivers were invoked in 94% of reviewed cases without subsequent public disclosure, contradicting stated transparency commitments. The researchers noted this finding is based on a sample of contracts that entered public records — contracts that never appeared publicly cannot be assessed.

[Response section — journalist to complete. Three parties have been contacted: the responsible minister's office, the contracting company, and the Procurement Authority. Responses due by 16:00. Include responses verbatim or in full summary as received.]
Sources · 5 verified
leaked_contract_v3.pdf
Pages 3, 18, 41 · Authenticity verified
94% confidence
Companies House · CH-2024-4892
Primary register · Verified match
Primary source
procurement.gov.uk archive
3 archive dates · No matching record
Confirmed absent
Government spokesperson
Emergency waiver claimed · Unverifiable
Disputed
KCL · Procurement Review 2024
Peer-reviewed · Core finding · p.341
Academic
Awaiting response
⏳ Minister's office
Contacted 10:22 · Deadline 16:00
⏳ Contracting company
Contacted 10:35 · Deadline 16:00
✓ Procurement Authority
Declined to comment · 11:48
AI disclosure
VERITAS GraphRAG used for claim verification, OSINT enrichment, and document cross-referencing. All AI findings reviewed by journalist. Methodology available at [link]. Published 1 June 2026.