Evidence and verdicts

Every result the Data API returns has been checked against each of your requirements. The result lists those requirements one by one, with the evidence that each is true.

Three verdicts

Checking one requirement against one ends in exactly one verdict.

matched

Evidence confirms the requirement is true.

not_matched

Evidence shows the requirement is false.

unknown

Evidence is missing, or too imprecise to decide.

The unknown verdict keeps guesses out of your results:

Missing data is not a “no”. A company with no published revenue is not “under $50M”. It is unknown.
Unknown is not a “yes”. A candidate with an unknown requirement is never returned as a match.
An imprecise figure is not rounded in your favor. A company known to have 51 to 200 employees is unknown for “more than 100 employees”.

What gets returned

A candidate is added to the results only when every requirement is matched. Here is a search with three requirements, checked against three candidates: So every entry in a result’s matched_criteria has verdict: "matched". The other outcomes are counted in the search’s progress:

{ "progress": { "candidates": 40, "accepted": 6, "rejected": 31, "unknown": 3, "not_returned": 0, "contacts_found": 0 } }

A high unknown count means matches may exist that could not be confirmed. Loosening the requirement that is hardest to verify, or replacing a free-text criterion with a structured filter, can turn unknowns into results.

Reading `matched_criteria`

Each result has one entry per requirement of the search. This example shows three:

matched_criteria (shortened)

{
  "matched_criteria": [
    {
      "criterion": "company.headquarters.state",
      "verdict": "matched",
      "evidence": { "summary": "Headquarters state: Texas", "sources": [], "items": [] }
    },
    {
      "criterion": "open_jobs",
      "verdict": "matched",
      "evidence": {
        "summary": "1 open job matches.",
        "sources": ["https://lakeviewdental.example/careers/sdr"],
        "items": [
          {
            "type": "job",
            "title": "Sales Development Representative",
            "posted_at": "2026-10-29",
            "location": "Austin, TX",
            "url": "https://lakeviewdental.example/careers/sdr"
          }
        ]
      }
    },
    {
      "criterion": "criteria.0",
      "verdict": "matched",
      "evidence": {
        "summary": "\"Family-owned and independently operated since 1998.\"",
        "sources": ["https://lakeviewdental.example/about"],
        "items": []
      }
    }
  ]
}

`criterion`

Identifies which requirement the entry is about:

Value	Refers to
A field name, such as `company.industry`	The filter condition on that field
`open_jobs` or `new_hires`	The condition on the company’s job listings or recent hires
`criteria.0`, `criteria.1`, …	The free-text criterion at that position in `criteria`, counting from zero

For a search started with query, the filters and criteria are the ones in the search’s interpretation.

`evidence`

string

A one-line explanation, or a direct quote from the source in quotation marks.

string[]

URLs that support the verdict. Empty when the evidence is a data record with no public page, such as an employee-count range.

object[]

The supporting records, when the requirement is about jobs, hires or technologies. Each has a type of job, hire or technology and the fields for that type.

What counts as evidence

Evidence has to be about the right company and has to state the thing you asked for.

Requirement	Counts	Does not count
”Offers same-day dental implants”	The clinic’s own page: “We provide same-day implant procedures”	A software vendor’s page: “We help clinics offer same-day implants"
"Is hiring SDRs”	An open SDR job at the company	A blog post from last year about growing the sales team
”Uses Salesforce”	A detected Salesforce deployment	A job description that lists Salesforce experience as nice to have
”Opened a new factory”	An announcement that the factory opened	An announcement that construction is planned

An AI model’s own reasoning is never evidence. If a claim cannot be tied to a source or a data record, the verdict is unknown.

Using evidence in your product

Show evidence.summary and the first source next to each result, so your users can check the list for themselves.
Pass sources through to an AI agent, so it can cite where a claim came from.
Store matched_criteria with the record. It is the reason the record entered your pipeline, which helps when someone audits the list later.

Evidence reflects what was true and findable when the search ran. A job listing can close and a page can change. found_at on each result records when it was added.

Get started

Concepts

Capabilities

Guides

Three verdicts

matched

not_matched

unknown

What gets returned

Reading `matched_criteria`

`criterion`

`evidence`

What counts as evidence

Using evidence in your product

Next steps

Search lifecycle

Result fields

​Three verdicts

matched

not_matched

unknown

​What gets returned

​Reading matched_criteria

​criterion

​evidence

​What counts as evidence

​Using evidence in your product

​Next steps

Search lifecycle

Result fields

Three verdicts

What gets returned

Reading `matched_criteria`

`criterion`

`evidence`

What counts as evidence

Using evidence in your product

Next steps