Where it helps
A burst of near-identical applications arrives for a remote role. The agent flags several as suspicious, and a recruiter reviews each summary before deciding what to do.
My take
MY PRACTITIONER VIEW
Fake and bot applications are a real problem, so this is welcome. But "suspicious" is a label a real person will carry, so agree how recruiters check and clear it before you switch it on.
What it won't do
- Oracle says it doesn't automatically reject, advance, rank or otherwise decide candidate outcomes.
- Once a candidate is marked suspicious, it doesn't rerun for them, and cleared statuses aren't re-evaluated automatically.
- It re-evaluates only when a profile is updated or the last evaluation is more than five years old.
- Oracle doesn't list the specific signals it uses.
Good first agent?
No. It labels real people, so it needs your most mature review process.
Risk level: Higher. It labels real people as suspicious, which can shape how they're treated, even though it doesn't decide outcomes.
Before you switch it on
- Who reviews flags, and how quickly?
- What do you need to see before acting on a flag, and how is clearing one recorded?
- Have your legal and data protection teams reviewed it, including how you'd explain a flag to a candidate?
Setup at a glance
Open Setup at a glance for the setup steps. Select the agent's type label to see what that type means.
Test it after each update
For Oracle Fusion Applications, non-production (test) quarterly updates follow the weekend after the first Friday of the update month; production follows the weekend after the third Friday. Run these checks after the test update and before production, and verify your tenant's exact dates in Oracle update notifications.
- Run it on test candidates and have a person review every flag.
- Clear a flag and check the profile and applications update together.
- Check that users without the clear privilege can't clear one.
Watch out
Turning it off hides its screens, including the Authenticity filter, but keeps existing data. Don't confuse it with the confidence score in the Job Applicant Screening Agent (26A): this is a separate agent with its own setup.
How it's changed
26D arrived.
- 26DIdentify potentially suspicious candidates using an AI agentOracle's note ↗
Related agents
Last reviewed November 2026
This is an independent practitioner view, not Oracle documentation. Features, names and availability change each release and depend on your licences and setup; some agentic applications carry extra costs. Always check Oracle's own notes and your environment.

