No score
Every result is a checkable fact with a source, never a number or a risk band.
No auto-reject
Candora never decides. GDPR Article 22 is a design constraint, not a policy page.
EU by construction
Database, storage and processing in an EU region. No third party outside the EU in the data path.
Auditable by default
Every finding carries its source, the time of the query and the rule version.
The problem
Screening was built for a world where the CV was roughly true.
Fabricated employment history is cheap to produce and expensive to catch. Today it gets caught at the end of the funnel, by a background check on the one candidate who already got an offer, or it does not get caught at all.
37%
of candidates who misrepresent their history do it on dates and duration of employment
Employment screening research, 2025
1 in 4
candidate profiles worldwide will be fake by 2028
Gartner, 2025
38.5%
of candidates flagged in AI monitored interviews
Fabric, 2026
1
candidate per hire gets a background check, out of hundreds screened
Standard hiring practice
Numbers are cited with their source and date, and we update them when the source does. If a claim on this site has no source next to it, treat it as an opinion.
How it works
From application to decision in five steps.
Candora sits at the top of the funnel, where the volume is, not at the end where the offer already went out.
- 01
The application arrives
A CV comes in through your ATS, by email, or by upload. The file is parsed in memory and never written to disk.
Your ATS or an upload
- 02
Claims are extracted
Employer, role, dates, contact details and document metadata become separate, checkable claims. Special categories under GDPR Article 9 are dropped at this step, even when the CV contains them.
Extraction model hosted in the EU
- 03
Each claim is checked against a source
Company registries, domain registration records, web archives, mail and phone infrastructure. Only company and domain data leaves Candora. The candidate's name never does.
ARES, EU registries, RDAP, web archive
- 04
The evidence report is written
Findings, verified claims and unverifiable claims are kept in three separate blocks. Every finding carries the source, the time of the query, the rule version and a sentence a person can read.
Candora report
- 05
A person decides
The recruiter reviews, agrees or overrides, and the decision is logged. Candora has no reject path, not in the interface and not in the API.
Your team
The report
Three answers. No fourth one.
Most tools in this category return a risk score. A score merges what was checked with what could not be checked and hides the difference. Candora keeps them apart, because merging them means claiming something we did not find.
Something does not line up
A claim contradicts a source. The report states the contradiction, names the register it came from, and shows the time of the query. It is a question for the recruiter, not a verdict on the candidate.
Findings
The company was registered 14 months after the candidate says they started there. Source: ARES, queried 4 August 2026.
Why not a score
A score is a probability. A probability cannot be shown to a candidate who asks why, cannot be defended to a regulator, and cannot be audited. A finding with a source can be all three.
Signals
Every rule states a fact you can go and check yourself.
These are the actual rules running in the product, with the source each one reads. Nothing here is inferred from writing style, nationality, language or behaviour.
ARES and EU business registries
Business registry
employer.not_found_in_registryThe employer named on the CV is not in the register that covers it.
employer.founded_after_claim_startThe company was registered after the date the candidate says they started there.
employer.dissolved_before_claim_endThe company was dissolved before the date the candidate says they left.
RDAP
Domain registration
domain.registered_after_claim_startThe employer's domain was registered after the claimed start date.
domain.younger_than_thresholdThe domain is younger than the threshold set for this ruleset version.
Public web archive
Web archive
domain.no_archive_snapshot_in_periodNo archived snapshot of the employer's site exists inside the claimed period.
The CV itself
Internal consistency
cv.period_reversedAn end date falls before its own start date.
cv.claim_period_in_futureA claimed period lies in the future.
cv.overlapping_employmentTwo full time roles overlap in time.
The CV file
Document metadata
cv.created_after_application_receivedThe document was created after the application was received.
cv.modified_before_createdThe document reports a modification date earlier than its creation date.
DNS and public disposable domain lists
Mail infrastructure
email.domain_has_no_mxThe email domain cannot receive mail.
email.domain_is_disposableThe email domain is a known disposable mail provider.
email.domain_younger_than_thresholdThe email domain is younger than the threshold.
Two teams, one report
The person who feels the problem and the person who owns the risk.
Talent acquisition
Check every application without slowing down and without being unfair.
Blunt filters on language, nationality or CV format are both discriminatory and blind. Candora gives you a specific reason to look twice at a specific application, and nothing to act on for the rest.
- Runs at the top of the funnel, on every application, not on the one finalist
- A thin digital footprint is never a finding
- Every flag comes with a sentence you can read out loud to the candidate
Security and compliance
A hiring control you can put in front of an auditor.
Organised infiltration uses real stolen identities that pass every check asking whether a real person exists. What does not survive is the employment history built around that identity.
- Append only event log, every event carries the rule version in force
- No prohibited practice under the AI Act: no emotion, deception, biometric or behavioural inference
- EU hosting, row level tenant isolation, retention you configure
Product boundaries
The list of things we will not build.
Most of this category ships features that are already unlawful in the European Union, or that have no evidence they work. We wrote our limits down before we wrote the code, and they are enforced by tests that fail when the limit is crossed.
No AI text detection
61% false positive rate on writers whose first language is not English. It punishes the wrong people.
No emotion or deception detection
A prohibited practice under Article 5 of the AI Act. It is also not supported by evidence.
No biometric template
The moment a template exists, the processing falls under Article 9 of the GDPR, where hiring has no usable legal basis.
No software on the candidate's device
No screen, process, clipboard or camera monitoring. It is intrusive, it is contested in several member states, and a second device defeats it anyway.
No gaze, keystroke or behaviour tracking
No construct validity study behind it, and documented bias by skin tone in the tools that use it.
No enrichment from third party data brokers
Article 14 of the GDPR requires telling the candidate what was collected and where from. The regulator has already fined this practice.
Try it
Open a real evidence report right now.
Three sample candidates, the same rules that run in the product, the same three blocks. It runs in your browser. We do not ask for an email and there is nothing to install.
- No signup, no email, no cookie wall
- Sample candidates, no real personal data
- Takes about ninety seconds
Three employers, all Czech, all inside covered registries.
Two findings, one of them a direct contradiction with the register.
Mostly outside covered registries. This is what an honest gap looks like.
Questions
The things people ask first.
Is this a background check?
No. A background check is a records search on one person, run late, on the candidate who already has an offer, and it costs enough that you run it once. Candora runs at the top of the funnel on every application, reads different sources, and returns findings rather than a record extract. Many customers will want both.
Do you scrape LinkedIn?
No. Candora reads data the candidate or the recruiter provides, plus public registries and APIs that permit the use. Scraping a professional network for hiring decisions does not have a defensible legal basis in the European Union and we did not want the product built on one.
What happens to a candidate with almost no digital footprint?
Nothing. An absent footprint is never a finding. Claims we cannot check land in the third block of the report, marked as not verified, with the reason. That block is descriptive, not negative.
Can Candora reject a candidate automatically?
There is no reject path in the product. Not as an API route, not as a default, not as a pre-selected recommendation in the interface. This is enforced in the code and covered by a test that fails if anyone adds one.
What does the candidate see?
Whatever you decide to show them, and we make that easy rather than awkward. Every finding is written as a sentence a person can read, with the register it came from, so a recruiter can raise it in a call without accusing anyone.
Where does the data live?
In an EU region. Database, object storage and the functions that process requests. There is no third party outside the EU in the data path. The CV file itself is never written to disk: it is parsed from the request and the text is discarded after extraction.
Which ATS do you work with?
On day one, email and upload, so nothing has to be integrated before you can use it. Integrations follow the pilot, driven by what pilot customers actually run. Candora is a layer, not a replacement for your ATS.
See what a report actually says before you talk to anyone.
The demo is the product argument. If it does not convince you in ninety seconds, a sales call will not either.