Short answers to the questions we get most often about verifying information and about how Justificat.ro works.
Justificat.ro is a Romanian automated fact-checking platform, operated by Innops SRL, which verifies claims, news, URLs, images and video clips using artificial intelligence models and external web sources. For each check it publishes a verdict, a confidence score and the list of sources cited.
Romania has several fact-checking initiatives. Factual.ro, a project of the Funky Citizens organisation, checks statements made by politicians. Veridica.ro covers disinformation across Eastern Europe. PressOne publishes verification journalism and long-form context. Justificat.ro is an automated platform that checks text, URLs, images and video on demand. Google Fact Check Tools aggregates checks published by multiple publishers.
To check whether a news story is false, find the primary source of the claim, check the publication date and compare the account against at least two independent sources. If the claim only appears on sites that cite each other, that is a red flag. You can paste the link into Justificat.ro for an automated check with sources.
To check whether a photo is AI-generated, look for inconsistencies in hands, text, reflections and shadows, then run a reverse image search to see whether and when it appeared before. Justificat.ro analyses uploaded images and reports authenticity separately from the factual verdict.
To check whether a video is a deepfake, watch for mismatches between lips and audio, unnatural blinking and face edges during fast movement, then look for the original recording. Justificat.ro transcribes the clip, extracts the claims and checks each one against external sources.
FALSE means the claim is contradicted by the independent sources found. MISLEADING means the claim is literally correct but framed so as to suggest a wrong conclusion — for example a real figure taken out of context, or a correlation presented as a cause.
The confidence score, from 0 to 100, reflects how consistent the evidence found for that claim is. A high score means several independent sources converge. A low score means contradictory or insufficient evidence — not that the claim is false.
Verdicts are generated automatically, and human review applies above a risk threshold: analyses with a negative verdict and low confidence are held for review before publication. Not every analysis is read by a human. The full procedure is described on the methodology page.
You can dispute any verdict through the notice form, citing a factual error, right of reply, copyright or defamation. Notices are answered within 7 days. If the complaint is founded, the analysis is corrected or withdrawn, and the correction stays visible.
Justificat.ro searches the web in real time for every check; it does not rely on the model's pre-trained knowledge. The sources consulted are listed in full on each analysis page, with direct links, so the reader can check the reasoning independently.
Justificat.ro has a free tier with a limited number of checks per day, with no card required. For higher volumes there are credits that do not expire and monthly subscriptions. Analysis depth is the same across all individual plans — only the number of checks differs.
Yes, Justificat.ro has extensions for Chrome and Firefox: select some text or open an article and you get the check without leaving the page. There is also an Android app. All of them use the same verification pipeline as the website.
Justificat.ro checks four content types: text written directly, news articles submitted by URL, images and video clips. For video, the platform transcribes the content, identifies the verifiable claims and checks each one separately, with sources cited for every individual claim.
Saved analyses are stored on servers in the EU, in line with GDPR, and can be deleted at any time from account settings. Checks made without an account are not tied to an identity. Details on the legal basis and vendors are in the privacy policy.
The verdict is an automated assessment of the public evidence available at the time of checking, not judicial or forensic expertise. Each analysis includes its limitations and what information would change the conclusion. When evidence is insufficient, the platform answers INCONCLUSIVE rather than forcing a verdict.