AI Search

Ask for the person you need. See why the results fit.

A recruiter’s question becomes filters you can read and edit, and every result carries the evidence it matched on, including what is missing.

Search
A search reading "payroll officer in Sydney, available within 4 weeks". Above the results, the filters it produced and the one word it could not use. Beside each person, the evidence for the match, when they were last spoken to, and how much of the question they answer.
Search in Recruited, running the question above it. The workspace is a demonstration one and every name in it is fictitious.

What changes

  • The search you meant

    You see the filters before anything runs, so a misread question is corrected in two seconds rather than discovered three shortlists later.

  • A shortlist you can defend

    Every result comes with the evidence and the date behind it, which is what a client asks for when they push back.

  • Nothing quietly excluded

    A candidate missing a field is shown with the gap named, not dropped for a reason nobody can see.

Start to finish

One role, from question to shortlist.

The worked example is a real recruitment problem: a part-time senior bookkeeper in Brisbane with Xero.

  1. Ask it the way you would say it

    “Find a senior bookkeeper in Brisbane who has Xero experience and is open to four days a week.” No syntax, no field names, no boolean.

  2. Read the plan

    Occupation, seniority, location and radius, skills, work pattern, freshness and source scope, each with the words it came from. This is the moment to fix “Brisbane” meaning 25km rather than the whole of South East Queensland.

  3. Adjust and run

    Widen the radius, drop the seniority, add a second accounting package. The filters are the query; changing them is the search.

  4. Read the reasons

    Each result explains itself: what matched, from where, and when it was last confirmed. Uncertainty is stated rather than hidden inside a score.

  5. Save it

    A search you will run again becomes a saved search or a pool, so next quarter’s version of this role starts from where this one finished.

The contract

What a result has to carry before it is allowed to rank.

A score with no reasoning looks authoritative and cannot be argued with. A result that cannot show its work does not get to sit at the top of the list.

  • Evidence

    The field or document the match came from, so it can be checked in one click.

  • Recency

    When that evidence was recorded. A skill confirmed in 2019 is not the same claim as one confirmed in June.

  • What is missing

    Named explicitly. “Work pattern not recorded” is more useful than a silently lower rank.

  • Provenance

    Which source scope produced it, so a result from a widened scope is never mistaken for one of your own records.

Controls

What it will not do.

The limits are the product.

  • No hiring decisions

    It ranks and explains. A person shortlists, submits and decides.

  • No outreach

    Finding someone and contacting them are separate actions, and the second one is a person’s.

  • No protected-trait inference

    Not inferred, not scored, not filterable, not displayed.

  • No unbounded sources

    Source scope is explicit on every search and defaults to your own records.

Questions

Does it crawl the web or LinkedIn?

No. It does not crawl anything, and it does not read a site’s members-only area. Public market discovery is a separate capability, Opportunity Finder, which keeps a public observation strictly apart from a candidate record.

Can I see and change what it understood?

Yes, before it runs. The question becomes a set of named filters with the words each one came from. You edit them and the search changes with them.

What does “why this matches” show?

The evidence: which field or document the match came from, and when it was recorded. Where something is missing or stale, it says so rather than quietly ranking the candidate lower.

Does it infer anything about a person?

No protected characteristic is inferred, scored, filtered on or displayed. Where seniority or a similar judgement is inferred from scope rather than stated, the result says it was inferred.

Does it contact anyone?

No. Search finds people. Contacting them is a separate, deliberate action by a person, under the consent and preference controls on the record.

Is my data used to train a model?

Not across customers. Retrieval is scoped to what your workspace can see, and your data is not used to train models shared with other customers.

Ready when you are

See Recruited on one of your live roles.

A demo against a real role tells you more than a feature list. Bring one, and we will walk it through the platform with you.

We reply within one Australian business day.