AI transparency

Classification

Vianora computes rule-based: fixed, disclosed formulas in database functions and in the program code. There is no learning model, no rule inferred from data and no language model. In our own assessment this is therefore not an AI system within the meaning of Art. 3(1) of Regulation (EU) 2024/1689 (AI Act).

Annex III No. 4 of the AI Act covers systems for recruitment, selection and decisions about individual workers. Vianora evaluates markets, occupations and regions and does not assess any individual person. In our assessment Annex III No. 4 therefore does not apply. The transparency and oversight rules are nevertheless applied here voluntarily.

No self-learning model and no language model is used. All evaluations rest on disclosed calculation rules documented under Methodology.

Data sources

Official labour market statistics of the Federal Employment Agency: reported vacancies, jobseekers, vacancy duration, jobseeker-to-vacancy ratio.

Publicly available job adverts and their history: time on market, reposting, likely stale listings.

Legal access rules: recognition, licence to practise, chamber membership, language level, visa and salary thresholds.

Details entered by users in the check: target occupation, target region, qualification, country of qualification, work experience, language level.

Weighted criteria

Ratio of jobseekers to reported vacancies in the occupation and district.

Vacancy duration compared to the national median.

Legal hurdles and their duration in months.

Pay level compared to statutory thresholds and the district median.

Quality and age of the underlying data. Unverified occupations are never shown as open.

Excluded characteristics

The following are explicitly never used as criteria: ethnic origin, gender, religion or belief, disability, age and sexual identity (Section 1 AGG, German Equal Treatment Act).

There is also no indirect assessment through proxies such as name, photo, personal nationality, marital status or residential postcode. The country of qualification is used solely for the legal recognition question and the visa route, never to assess a person.

No automated selection of applicants

Vianora does not rank, filter or screen out people. Companies only receive market and occupation analyses plus non-binding suggestion lists without score-based exclusion.

Rejection, pre-selection or scoring of individual applicants does not take place in the system and is not intended.

Human review and objection

Every evaluation can be reviewed. A short message to info@vianora.eu stating the occupation, the district and the date is sufficient; the review is carried out by a person, not by the system.

In addition, the rights to access, rectification, erasure, restriction, data portability and objection under Art. 15 to 21 GDPR apply. Profile, watchlist and account can be exported and deleted in your own area.

Incorrect or outdated data is corrected or removed from display once reported.

Limits

The evaluation describes market and legal situations, not a person's suitability. It does not replace legal, migration or career advice and makes no decision in an individual case.

How the evaluations are computed

The traffic light is a database function with a fixed order: first hard access barriers and visa status, then the vacancy period or, failing that, the observed time online against a fixed threshold, finally the jobseeker-to-vacancy ratio against a fixed limit. All thresholds are stored in the constants table and are shown under Methodology. Identical inputs always yield the same result.

The review cascade works through six defined stages in sequence and stops at the first stage that is not met. It does not weight and does not learn; it checks conditions.

The indicators are sums, medians and ratios over official and collected values. Missing values are not estimated by a model: if a value is missing, the gap is stated.

In our assessment the traffic light, the review cascade and the indicators are therefore rule-based calculations and not AI systems within the meaning of Art. 3(1) of the AI Act. On that assessment Art. 50 does not apply to them.

Machine-generated text

No language model generates text in the running application. Verdicts, traffic-light reasons, cascade statements and data-status notes are assembled from fixed text blocks into which computed values are inserted.

The editorial texts of the application, including methodology, legal texts and this page, were drafted with tool support and then reviewed and taken responsibility for by the operator. Whether such texts count as informing the public on matters of public interest within the meaning of Art. 50(4) is not clear-cut; the labour-market evaluation texts are the most likely candidates.

Rule for further development: if a language model is integrated in future, every output generated or substantially edited by it will be labelled as machine-generated at the place where it is shown. If a chatbot is added, the interaction must be recognisable as machine interaction before the first exchange.

Legal status and deadlines

Regulation (EU) 2024/1689 entered into force on 1 August 2024 and applies in stages. The prohibitions under Art. 5 have applied since 2 February 2025; general applicability, including the transparency duties of Art. 50, since 2 August 2026.

Under the so-called Digital Omnibus, the obligations for stand-alone high-risk systems under Annex III were, according to reports, postponed to 2 December 2027. Status of this statement: 2 August 2026. The exact wording of the amending regulation still has to be verified.

Art. 50 requires labelling for direct interaction with an AI system, emotion recognition, biometric categorisation and synthetically generated content. Vianora runs no chatbot, recognises no emotions, performs no biometric categorisation and generates no synthetic content. If a language model is added later, every output produced with it will be labelled as machine-generated where it is displayed.

The competent contact and complaints body in Germany is the Bundesnetzagentur, designated as the central authority for the AI Act. Complaints about algorithmic evaluation in this application can be raised there and, in addition, with the competent data protection authority.

For content placed on the market before 2 August 2026, the transitional rule allows until 2 December 2026. Existing texts in this application will be reviewed for labelling requirements by that date.