Content input
Advertisement, claim or message submitted for review.
An innovation project in development
Verify before you publish.
Validni is developing an AI-assisted platform to help businesses assess marketing claims and advertising content before publication — with a focus on evidence, relevant regulations and authoritative sources.
Currently in the concept and development-planning stage
The proposed workflow
From marketing claim to evidence.
The planned platform is designed to organise a preventive review into clear, traceable steps. This workflow describes the intended product concept, not a currently operating service.
AI-assisted review
Identify claims and potential risk indicators.
Source checking
Compare with relevant rules and authoritative information sources.
Evidence report
Structured findings, references, evidence trail and limitations.
The product scope
15 compliance areas. One preventive workflow.
The planned catalogue contains eight core areas and seven expansion areas. The Serbian MVP will validate a smaller set of priority use cases first.
Catalogue scope is a long-term product direction, not a claim that all 15 areas are already implemented.
Green Claims
Environmental and sustainability claims, evidence and substantiation.
Prices & Discounts
Price presentation, discounts, reference prices and promotional clarity.
AI Transparency
Transparency around AI-generated or AI-manipulated content.
Dark Patterns
Manipulative interface patterns and consumer choice architecture.
Health Claims
Health-related marketing claims and their supporting evidence.
Hidden Advertising
Disclosure of sponsored, influencer and commercial communications.
PDO / PGI Protection
Use of protected designations of origin and geographical indications.
Fake Reviews
Review authenticity signals, misleading ratings and testimonials.
Expansion 09–15 · Planned additional areas
Guarantees & Conformity
Consumer guarantees, conformity of goods and related information duties.
Subscriptions & Cancellation
Recurring payments, renewal terms, cancellation and withdrawal flows.
Financial & Investment Claims
Marketing statements about financial products, returns, risks and investments.
Comparative Advertising
Comparisons with competitors, products, prices and performance claims.
Marketing & Legal Consistency
Consistency between public marketing messages, terms, policies and supporting documents.
Children & Vulnerable Groups
Advertising practices and claims directed at children or vulnerable audiences.
Product Information & Declarations
Product descriptions, mandatory information, labels and declarations.
Module coverage will be introduced and validated progressively for each market. The catalogue does not represent completed functionality.
Implementation strategy
A phased path from Serbia to wider markets.
Prioritise practical value for Serbia first, then expand compliance coverage and geographic reach after MVP validation. MVP priority is not the same as the full module catalogue.
Serbia-first MVP
Build and validate a focused product with reliable Serbian sources and a usable review workflow.
- Prices and discounts
- Hidden advertising
- Fake reviews
- Guarantees and conformity
- Subscriptions and cancellation
Deeper compliance coverage
Extend the module catalogue, evidence handling and cross-checking of commercial content.
- Green and health claims
- Financial and investment claims
- Dark patterns
- Marketing and legal consistency
- Comparative advertising
Regional and EU readiness
Add jurisdiction-specific rules and trusted data providers after the Serbian MVP demonstrates value.
- Children and vulnerable-group advertising
- Product declarations and information
- AI transparency
- PDO / PGI protection
- Country-specific source integrations
Indicative sequencing, not a delivery-date commitment. Exact release scope depends on project validation, source availability and regulatory review.
Beyond business compliance
A practical information and protection platform for consumers.
Validni has the potential to grow beyond a business-facing compliance tool into a trusted public resource: helping people understand commercial messages, recognise warning signs, find reliable information and make more informed purchasing decisions.
Understand your rights
Plain-language guides to consumer rights, guarantees, conformity, returns, subscriptions, cancellations and complaints — linked to relevant rules and official institutions.
See through advertising claims
Explanations of price reductions, environmental and health claims, sponsored content and influencer promotions — showing what is evidenced, unclear or still needs checking.
Check products and offers
A planned checker for a product page, advertisement or offer, with relevant claims, available evidence, potential red flags and source links.
Recognise suspicious patterns
Educational examples of unverifiable reviews, hidden sponsorship, misleading countdowns, pressure tactics, unclear recurring payments and difficult cancellation flows.
Find trusted alerts and sources
A future information hub for official product-safety notices, recalls, consumer alerts, public registers and institutional guidance, with dates and original links.
Know where to turn next
Step-by-step signposting to the seller, market-surveillance body, consumer organisation or official complaint channel. Validni would not adjudicate disputes.
Evidence, not unsupported ratingsSeparate verified facts, source statements, user reports and AI-generated indications.
Accessible to ordinary peoplePlain language, practical examples and mobile-friendly access.
Responsible handling of reportsFuture community reporting would require moderation, privacy safeguards and a right-of-reply process.
Consumer-facing features are a proposed development direction, not currently available services. Information would be dated and sourced. Automated signals alone would not establish a legal violation.
Technology & evidence infrastructure
AI can analyse a claim. Authoritative sources provide the basis for verification.
Validni is planned as a provider-independent, modular platform. AI services support research and analysis, while a separate national and regulatory source layer anchors findings in traceable information.
Architecture shown is a planned design direction, not a statement that integrations or services are already operational.
01 · USER INPUTAd · offer · claim · document
↓
02 · SUPERVISOR & ORCHESTRATIONTask routing and controls
↓
AI · EXECUTION LAYER
AI & provider pool
Model APIs, routing, quotas, fallbacks and cost controls.
AG · INTELLIGENCE LAYER
Agents & Skills
Specialised agents and validated reusable procedures.
✓ · EVIDENCE LAYER
Data Providers & Sources of Truth
Official legislation, national agencies, regulators, public registers and trusted institutional platforms.
↓
03–04 · EVIDENCE & STRUCTURED REPORTFindings · sources · limitations
Planned infrastructure controls
Agent RegistrySkill RegistryCapability RegistryTest RegistryIntegrity & permissionsFree-tier options where available
Under the hood · planned system design
From an input to an evidence-backed review.
The following blueprint explains the intended processing pipeline, control points and evidence model. It is a development specification for the proposed platform, not a description of deployed functionality.
Structured content intake
A business user submits an advertisement, product page, offer, document, image or text. The intake step records the selected market, language, date, product category and intended audience, then prepares the material for analysis.
Task planning and orchestration
A supervisor converts the request into a controlled workflow: identifies the jurisdiction, classifies likely compliance areas, selects approved agents and tools, and defines the evidence needed. It manages workflow state, checkpoints, retries and hand-offs.
Provider-independent model pool
A model router assigns suitable providers to extraction, classification, summarisation or reasoning tasks. Planned controls include provider health checks, usage quotas, cost ceilings and fallback routes. Free tiers may be used where available; availability and terms can change, so no provider is assumed to be permanently free.
Specialist agents and versioned skills
Focused agents handle defined areas such as prices and discounts, hidden advertising, fake reviews, guarantees or subscriptions. Skills are reusable procedures with defined input/output schemas, versions, integrity checks and test history. An agent cannot grant itself new permissions.
Jurisdiction-aware source retrieval
A curated knowledge layer retrieves relevant legislation, regulator guidance, public registers and institutional materials. Each source is tagged by country, issuing body, language, document type, publication date and effective period. Search snippets alone are not treated as legal evidence.
Deterministic checks plus AI analysis
Where possible, exact values are checked with deterministic logic: dates, percentages, arithmetic, required fields and document versions. AI supports interpretation of context and language. Ambiguous or high-impact cases are flagged for human review rather than converted into automatic legal conclusions.
Governance layer
Four registries keep the system organised and controlled.
Registries are the control plane for what can run, what it can access and how its quality is verified.
01 · Agent RegistryAgent identity, scope, version, permissions and approval status.
02 · Skill RegistryReusable procedures, schemas, integrity hashes and validation status.
03 · Capability RegistryExplicitly allowed tools and operations, with boundaries and conditions.
04 · Test RegistryFunctional, regression, adversarial and source-freshness test records.
Evidence ledger
Every finding should have a traceable basis.
The planned evidence record links each reviewed claim to the material used to assess it. This makes it possible to inspect not only the result, but also where it came from and what remains unknown.
Claim / claim excerptSource ID & URLIssuing authorityJurisdictionPublished / effective dateLast checkedSource versionEvidence gapConfidence & uncertainty
The report distinguishes observed content, official source statements, user-supplied material, AI-generated interpretation and unresolved questions.
Illustrative workflow
Example: an advertisement says “50% off”.
- CaptureRead the advertised price, discount claim, product and visible time period.
- RouteSelect the Serbian prices-and-discounts workflow and its approved checks.
- RetrieveFind the relevant official legal text and guidance, checking jurisdiction and effective dates.
- Compare evidenceCompare the claim with the reference-price history and documents actually supplied.
- Report limitsIf historical price data is missing, say “cannot verify from submitted material” and request evidence; do not invent a conclusion.
Illustrative report outputClaim detected · Applicable source located · Reference-price evidence missing · Further documentation required · Human review may be appropriate
Controls built into the design
- Role-based access and separation of business workspaces.
- Secrets management, encrypted connections and controlled data retention.
- Prompt-injection resistance and treatment of uploaded content as untrusted input.
- Audit events, rate limits, quotas, checkpoints and safe retries.
- Human escalation for ambiguous, incomplete or high-impact findings.
How the proposed system will be tested
- Accuracy of claim extraction and compliance-area classification.
- Relevance and correctness of cited sources.
- Jurisdiction and effective-date correctness.
- False-positive and false-negative analysis.
- Reproducibility, response time, provider availability and operating cost.
Implementation boundary: these components describe the proposed target architecture. The current website is a project presentation; it does not accept submissions, run compliance checks, maintain a live legal database or generate operational evidence reports. No institutional integration, data-sharing agreement or endorsement is claimed.
National data provider pools
One technical core. Market-specific rules and sources.
The platform can retain a common workflow while adapting legal mapping, language and official source providers to each jurisdiction.
Serbia · first market
National legislation, competent institutions, registers and official public information.
Phase 01 · MVPNeighbouring markets
Separate national source collections and regulatory mappings for each country.
Phase 02 · RegionalEuropean Union
EU legal sources combined with country-specific sources for relevant Member States.
Phase 03 · EUIllustrative Serbian source catalogue
Potential official sources, subject to relevance, authority, access terms and update-frequency checks:
Ministry of Internal and Foreign TradeLegal Information System of SerbiaNational Bank of SerbiaSerbian Business Registers AgencyCommission for Protection of CompetitionCompetent inspection portals
Illustrative EU source catalogue
Potential EU-level references, complemented by relevant national authority portals:
EUR-LexSafety GateEuropean Commission portalsNational competent authorities
These are potential official source providers only; they do not imply existing technical integration, data-sharing agreement or endorsement. Every source must be checked for jurisdiction, effective date, amendments and context. AI outputs are analytical assistance, not legal determinations.
Designed around trust
Evidence first. Clear limitations.
Validni is intended to make early review more structured, explainable and accessible — not to replace legal professionals or regulators.
01
Authoritative sources
Planned references to applicable legislation, official institutions and relevant source documents.
02
Traceable findings
A proposed evidence trail showing the basis, references and limitations behind each finding.
03
Preventive approach
Designed to help review content before publication, rather than only react after a dispute.
Starting in Serbia.
The first planned product scope focuses on the Serbian market. Regional and EU expansion would follow only after MVP development and market validation, using jurisdiction-specific rules and source providers.
Development status: Validni is currently at the concept and project-planning stage. The functional platform is proposed for development through the Smart Start project. No public product functionality is available on this website.