AI-generated content
The News section is our testing ground: articles, visuals, voiceovers, and videos are generated by an AI system and published automatically. We state this clearly because the law requires it. Besides, showcasing automation without disclosing it would be unprofessional.
What the system generates
HEXART AI Newsroom generates article copy, excerpts, subheadings, accompanying visuals, voiceover scripts, audio recordings, and short videos for selected stories. Topics come from a curated list, visitor questions approved in the admin panel, or automated queries for topics lacking good web answers. The model retrieves factual data via search while drafting.
Outside the News section, AI generated short background videos and selected visuals, and language versions of the site are translated automatically by an AI model. Descriptions of services, projects, and solutions are drafted with AI assistance using our internal documentation and implementation data. HEXART remains fully responsible for their content.
Automated pre-publication checks
Before publication, every piece undergoes two automated checks. The first evaluates copy: headline length and structure, category, paragraph layout, subheadings, and topic duplication against earlier posts. Copy that fails returns to the model with specific revision notes. After a third attempt, the system fills missing fields and publishes the strongest version.
The second check verifies completeness: at least two sources, rendered visuals, and finished voiceover audio. Incomplete materials are withheld. No human reviews the material prior to publication. As publisher, HEXART remains fully responsible for published content. Report errors to biuro@hexart.pl; we will correct the copy or remove the article.
These controls evolved over time and did not always perform as expected. We document five months of operational data, including failures, in our 224-article report.
These articles do not constitute legal, financial, or medical advice. Verify critical facts directly with primary sources before making business decisions. We include source links below articles only when they resolve directly to primary reference materials.
Content disclosures
Each article displays the disclosure "Tekst napisany z pomocą AI." Visuals carry the label "Obraz wygenerowany przez AI.", voice tracks state "Głos syntetyczny.", and videos show "Wideo wygenerowane przez AI." A disclosure note below the copy specifies which components were generated. Manually revised articles carry the label "Tekst napisany z pomocą AI i sprawdzony przez człowieka."
Separately, every piece includes machine-readable metadata. In the page structured data, we record the system as the technical creator alongside the value trainedAlgorithmicMedia from the IPTC Digital Source Type vocabulary. This allows search engines and aggregators to detect the origin of the content without scraping the page.
AI-generated videos and visuals elsewhere on the site display on-frame badges: "Wideo wygenerowane przez AI." for video, "Obraz wygenerowany przez AI." for visuals. Badges on visuals remain permanently visible. Video badges appear on initial load, collapse into an AI icon, and expand again on hover or tap.
Site assistant and plan generator
The site assistant in the corner runs on AI and identifies itself immediately. It answers using only hexart.pl content: relevant page excerpts are fed to the model, and responses cite the source URLs used. If the site does not contain the answer, the assistant states this plainly and offers a call with the founder. It does not extrapolate from general knowledge. Unanswered queries route to an admin queue, where a human decides whether to produce dedicated content.
The implementation plan generator operates in reverse compared to a chat: rules drawn from the site content select recommendations, stages, amounts, duties, and grant programs. Two identical forms will always generate an identical plan. The language model writes only the three-paragraph introduction and cannot add numbers or regulations not present in the plan. Each page of the plan indicates whether the introduction was generated by the model or produced by rules.
Legal basis
Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 (AI Act), Article 50(2) and (4). Paragraph 2 requires synthetic content to be marked in a machine-readable format and detectable as artificially generated. Paragraph 4 requires disclosure for deepfakes and for published text informing the public on matters of public interest. The exception for text covers material that has undergone human review and for which an entity holds editorial responsibility. Article 50 applies from 2 August 2026.
Questions and reports
An error in the material, doubts about a source, or questions about the mechanism? Write to biuro@hexart.pl. If you are interested in a similar process for your company, see details under services.
Frequently asked questions
Which content on the website is created with AI?
All articles in the News section, including illustrations, voice-over narrations, and short video clips. Other website copy is drafted with AI support based on our internal materials and implementation data. HEXART assumes editorial responsibility for this content.
Does anyone review these texts before publication?
No. Articles in the News section are published automatically. Before publication, the system checks text structure and content completeness: at least two sources, images, and a voice-over recording. Incomplete material is not published. HEXART assumes responsibility for published content as the publisher.
Where do the facts in the articles come from?
From a search engine integrated with the model. We publish source links below the article whenever the provider returns direct links to the source material. Older materials lack them: the tool used earlier returned redirect links that expired, so we removed them to avoid dead links.
How to report an error in AI-generated material?
Email biuro@hexart.pl with the URL of the article. Material with a confirmed error is either corrected or removed from the website.
Is AI disclosure required by law?
In certain cases. For content, Article 50 of the AI Act requires disclosure for deepfakes and for text informing the public on matters of public interest that has not undergone human review. The News section does not undergo human review, so we label it. We label the remaining content voluntarily. Article 50 applies from 2 August 2026.