Recruitment Media and Scout Services: Information Provision or Employment Placement?
Hello, this is Legal Agent.
When we advise on recruitment media, scout services, talent databases, matching platforms, and recruitment AI in Japan, one of the first questions is whether the service merely provides information or conducts an employment placement business.
A business may describe itself as a job board where employers and candidates communicate directly. That description alone is not conclusive. Depending on the actual product and operations, a licence for a fee-charging employment placement business may be required.
The question should therefore be addressed while the product is still being designed, not only when the terms of use are ready for launch.
Information provision or employment placement?
Under Japan's Employment Security Act, employment placement broadly means receiving job and job-seeking applications and arranging the establishment of an employment relationship between an employer and a job seeker.
Recruitment information provision, by contrast, covers specified forms of providing job information to prospective workers and providing job-seeker information to employers. If a service only provides information, does not receive job and job-seeking applications, and does not arrange employment, an employment placement licence is generally not required.
The Ministry of Health, Labour and Welfare's current guidance makes clear, however, that the legal characterisation depends on the substance of the service. In practice, we break the product down into at least five questions:
- Does the operator receive job applications or job-seeking applications from both sides?
- Does the employer or candidate choose what information to view, or does the operator independently decide who receives which information?
- Does the service only change display order, or does it add a new assessment such as an expected salary or probability of being hired?
- Do employers and candidates communicate directly, or does the operator edit, prioritise, delay, or otherwise control the communication?
- Do advertising, sales materials, contracts, and actual operations promise candidate introduction or successful hiring?
The assessment should cover the entire journey: registration, search, recommendation, application, messaging, interview coordination, offer terms, and charging. Looking only at a public search screen is rarely enough.
Automation does not remove the licensing question
The Ministry's guidance states that the relevant acts may require an employment placement licence whether they are performed manually or automatically by an electronic information-processing system.
Making all job information available and merely presenting it in a different order based on browsing history does not necessarily amount to legally relevant processing of the information. The assessment may change where the operator independently selects a limited set of jobs, excludes all other jobs, or adds personalised statements about salary or hiring prospects.
For recruitment AI, the important questions are not how sophisticated the model is but who determines the matching criteria, which candidates or jobs are excluded, and whether users can still access the underlying information.
Scout functions look different depending on their design
Many scout products allow an employer to search a candidate database and send a direct message. Providing a database and messaging function can look like information provision. The risk increases if the operator selects candidates, recommends particular individuals, edits communications, coordinates interviews, or negotiates employment conditions.
The fee model does not determine the legal classification on its own. It is nevertheless relevant to the overall substance. A service marketed as selecting the best candidate, operated through manual recommendations by customer-success staff, and paid only upon a successful hire warrants a closer review even if the terms say that the operator does not conduct employment placement.
For AI recommendations, identify whose criteria control the result
An employer may specify objective conditions such as a qualification, location, or years of experience. Delivering information in accordance with that request can be different from the operator creating its own score and deciding, without a specific instruction, which candidates the employer may see.
An apparently simple “recommended” label may be based on predicted application, hiring, attrition, or salary scores. Without understanding the score, training data, exclusion conditions, and scope for manual intervention, the business cannot fully assess licensing, discrimination, explainability, or erroneous recommendations.
Product documentation should identify input data, selection criteria, exclusion rules, recipients, the communication flow, and the changes an operator can make manually. A material model update should trigger a legal check where it changes selection, processing, or intervention in communications.
Candidate data and privacy
Recruitment media and scout services handle large amounts of candidate data: names, employment and education history, compensation, skills, job-change intentions, evaluations, interview records, résumés, portfolios, and social-media information. Some services may also receive information close to health data or reasons for leaving employment.
The business should analyse purposes of use, third-party provision, processors, joint use, retention, deletion requests, and security under Japan's Act on the Protection of Personal Information. A candidate should be able to understand which employer receives which information and at what stage.
Collecting information about prospective workers and providing it to employers may also constitute a specified recruitment information provision business that requires a notification. Concluding that the service is not employment placement does not remove all Employment Security Act obligations.
A candidate's decision to create a profile is not automatically consent to disclose all information to every employer. The product and privacy policy should agree on whether candidates control visibility, whether approval is obtained before identifying information is disclosed, and what remains after account deletion.
If recruitment AI infers job-change intent, aptitude, health-adjacent information, or other sensitive characteristics, the inferred data also requires governance. The service should limit what is provided to employers and preserve enough information to explain the basis of a recommendation.
Accurate presentation of recruitment information
Job information must also be accurate. Compensation, location, duties, employment status, contract term, probation, fixed overtime pay, remote-work conditions, benefits, and selection criteria can all cause disputes if the display differs from reality.
Terms should require employers to provide accurate information and keep it updated, but contractual wording may not be sufficient. The platform needs criteria and owners for investigating reports, suspending listings, correcting errors, preventing recurrence, setting update deadlines, and removing stale jobs.
AI-generated summaries create an additional risk because they may introduce wording the employer never supplied. Controls should prevent fixed overtime pay from being presented as base salary, an independent-contractor role from appearing to be employment, or an on-site requirement from being omitted. Important conditions may need fixed fields and review of differences between source and generated text.
The terms should address employer warranties, prohibited listings, update duties, review and suspension rights, candidate complaints, and allocation of responsibility. But stating “we do not conduct employment placement” does not override the actual service.
A legal review should therefore cover product screens, administrator functions, sales materials, pricing, operational manuals, and interviews with the teams that operate the service. Defining legal-review triggers in the development process makes it easier to identify a change in licensing or notification requirements before release.
Conclusion
Recruitment media and scout services should begin with the line between employment placement and recruitment information provision. The assessment turns on actual selection, recommendation, communication, interview coordination, negotiation, fees, and data flows—not the name attached to the service.
Recruitment AI can change those facts whenever the recommendation logic, pricing, or operating model changes. A practical starting point is to map the user journey, identify who makes each decision, determine whether users can access other information, and record any operator intervention in messages or selection.
LegalAgent advises recruitment media, scout products, recruitment AI, talent databases, and employment placement services on the Employment Security Act, privacy, product design, and terms for employers and candidates.