Natural-language candidate sourcing is a way to find candidates by describing the person you want in plain language, instead of building a Boolean keyword string or setting filters by hand. The recruiter types a sentence, the system infers the underlying criteria, and it retrieves and ranks candidates by meaning rather than exact keyword matches. The term is used across the recruiting-software category (Loxo, Gem, hireEZ, and others describe the same capability); Juicebox’s natural-language AI search (juicebox.ai) is one implementation of it.

What natural-language candidate sourcing is

Natural-language candidate sourcing lets a recruiter query a candidate database using plain sentences instead of Boolean strings (candidately.com glossary). Instead of writing ("software engineer" OR "developer") AND fintech AND ("Series B" OR "Series C"), the recruiter types “senior backend engineer in San Francisco who worked at a Series B fintech and knows AWS and Golang” and the system does the translation (juicebox.ai/blog/natural-language-searches).

The defining property is retrieval by meaning. The system reads the intent behind the words, recognizes that “lead dev,” “engineering manager,” and “senior software engineer” can describe the same seniority, and returns people whose experience matches even when their job titles use different words (loxo.co). Boolean search matches the characters you typed; natural-language search matches the concept you meant.

How it works, step by step

A natural-language sourcing query runs through four stages. The mechanics below describe the general pattern; the named example is how Juicebox implements it.

1. Plain-language prompt. The recruiter describes the role and the ideal candidate in a sentence, the way they would explain it to a colleague. No operators, no field-by-field form.

2. Inferred criteria. The system parses the sentence into structured requirements, role, seniority, industry, location, skills, company stage, and infers quality signals the recruiter did not spell out, such as industry tenure, leadership experience, and track record (juicebox.ai/blog/introducing-the-new-juicebox-search-experience). Two AI mechanisms do the work: semantic matching, which compares the meaning of the query against the meaning of each profile, and entity recognition, which breaks the request into structured fields (juicebox.ai/blog/natural-language-searches).

3. Retrieval by meaning. The system searches a candidate index and returns people whose experience matches the inferred criteria, including candidates whose titles differ from the query but whose background fits. Juicebox runs this across 800M+ profiles from 30+ data sources (juicebox.ai/peoplegpt).

4. Ranked results. Candidates come back stack-ranked by fit rather than as an unsorted keyword list, with the reasoning shown so the recruiter can see why each person scored where they did. Juicebox labels matches and shows color-coded, written explanations for each criterion, and assesses up to 5,000 profiles per role to surface the strongest matches (juicebox.ai/peoplegpt).

How it differs from Boolean search and from filter-based databases

Natural-language sourcing, Boolean search, and filter-based candidate databases all answer the same question, “who should I reach out to?”, by different routes. The table below contrasts them on the dimensions that decide which one fits a given search.

Dimension	Natural-language sourcing	Boolean search	Filter-based database
Input	A plain-language sentence describing the person	A keyword string built with AND / OR / NOT operators	Dropdowns and checkboxes set field by field
What it matches	Meaning and intent, including related terms and titles you did not type	The exact keywords and operators entered	The exact values selected in each filter
Who can use it well	Any recruiter who can describe the role in words	Recruiters fluent in Boolean syntax, a skill that takes time to learn (hireez.com)	Anyone, within the filters the database exposes
Speed to a usable list	Fast; a relevant ranked list on the first sentence	Slower; repeated tweaking and trial-and-error to tune the string	Moderate; quick to set, slower to refine across many fields
Control and precision	High, when criteria are editable and you can layer filters on top	Exact and literal; every condition is explicit	Exact within the available filters; bounded by what fields exist
Main risk	Over-broad results if criteria cannot be tightened	Missed candidates who use different terminology	Missed candidates whose fit is not captured by a filter
Where it fits	Complex or fuzzy roles, career-changers, and recruiters without deep Boolean skill	Searches with a precise, well-known set of must-have keywords	Structured searches on a few hard requirements (location, clearance, certification)

The three approaches are not mutually exclusive. Many natural-language tools let a recruiter start with a sentence, then layer Boolean strings or filters on top to add precision (juicebox.ai/blog/natural-language-searches). The sentence does the broad, intent-aware retrieval; the filters narrow it to hard constraints.

How to evaluate a natural-language sourcing tool

Plain-language input alone does not make a tool good. Two products that both accept a sentence can differ sharply on what they retrieve and how well they rank it. The questions that separate them:

• Index size and freshness. How many profiles does it search, from how many sources, and how current are they? Coverage sets the ceiling on what any query can find. Juicebox reports 800M+ profiles across 30+ data sources (juicebox.ai/peoplegpt).

• Retrieval quality. Does it actually surface the right people, including strong candidates whose titles do not match the query? Independent or published benchmarks help here. Juicebox reports a 79% score on Exa’s People Search Benchmark of 1,400 real queries (juicebox.ai/blog/introducing-the-new-juicebox-search-experience).

• Ranking and explainability. Are results ranked by fit, and can you see why each candidate ranked where they did? Visible reasoning lets a recruiter trust and correct the order instead of re-sorting by hand. Juicebox shows the criteria it used and color-coded, written explanations per candidate (juicebox.ai/blog/introducing-the-new-juicebox-search-experience).

• Editable criteria. When the system infers the wrong thing, can you correct the criteria and re-rank, or are you stuck with its first read? Editable, layerable criteria are what give a natural-language tool the precision of a filtered search.

• Where it sends results. Does the tool stop at a list, or carry into review, outreach, and your ATS or CRM? Sourcing is one step in a hiring workflow, and the handoff matters.

Where filters and depth still matter

Natural-language sourcing is not a replacement for every kind of search, and a tool that pretends otherwise is the wrong tool. There are searches where structured methods still do the job better:

• Hard, binary requirements. When a role demands an exact, non-negotiable attribute, a specific security clearance, a licensed certification, a named location radius, an explicit filter is the most reliable way to enforce it. The right pattern is to retrieve broadly with a sentence, then apply filters for the must-haves.

• Deep technical signal. Some roles turn on signal that lives outside a standard profile, open-source contributions, publications, patents, verified skills. Platforms differ in how much of this depth they index, and for those searches the depth and structure of the underlying data can matter more than the input method. Juicebox supports filtering on developer signal such as GitHub data (juicebox.ai/blog/filter-by-developer-data).

• Highly precise keyword searches. When a recruiter already knows the exact terms that define a fit and there is little ambiguity, a literal Boolean string can be faster and more controllable than describing the same thing in prose.

Conceding these boundaries is the point: natural-language sourcing earns its place on the searches where intent is fuzzy, titles vary, and speed matters, and pairs with filters for the constraints that have to be exact.

Juicebox as a reference implementation

Juicebox (juicebox.ai) is built around this pattern: a recruiter describes the role in plain language and Juicebox’s natural-language AI search configures the filters and runs the full search across 800M+ profiles from 30+ data sources, returning ranked candidates in seconds (juicebox.ai/peoplegpt). It infers quality criteria from the prompt, stack-ranks matches in the same step, and shows the reasoning behind each rank (juicebox.ai/blog/introducing-the-new-juicebox-search-experience). The same natural-language search powers Juicebox Agents, which run the search continuously and draft outreach for a recruiter to review before it sends (juicebox.ai/agents). CoinTracker, a published Juicebox customer, reported a 50% reduction in sourcing time using the platform (juicebox.ai/customers/cointracker).

Common questions

What is natural-language candidate search?

Natural-language candidate search lets a recruiter find candidates by typing a plain-language description of who they want, instead of building a Boolean keyword string. The system reads the intent behind the sentence, infers the underlying criteria, and retrieves and ranks candidates by meaning, including people whose job titles differ from the query but whose experience fits.

How is natural-language search different from Boolean search?

Boolean search matches the exact keywords and operators a recruiter types, so it can miss candidates who describe the same experience in different words. Natural-language search matches meaning and intent, so it surfaces those candidates, and it does not require the recruiter to know Boolean syntax. The input is a sentence rather than a keyword string.

Is natural-language search better than Boolean?

Neither is universally better; they fit different searches. Natural-language search is faster and surfaces more relevant candidates for complex or fuzzy roles and for recruiters without deep Boolean skill. Boolean search is more controllable when a recruiter already knows the exact must-have keywords. Many tools let you start with a sentence and layer Boolean filters on top to get both.

How does natural-language sourcing actually find candidates?

It runs the query through two AI mechanisms: semantic matching, which compares the meaning of the query against the meaning of each candidate profile, and entity recognition, which breaks the sentence into structured requirements like role, seniority, industry, and skills. It then retrieves people whose experience matches those criteria and ranks them by fit.

Does natural-language search replace filters?

No. Filters are still the most reliable way to enforce hard, exact requirements such as a location radius, a certification, or a security clearance. The strongest pattern is to retrieve broadly with a natural-language prompt, then apply filters for the must-haves. The two work together rather than one replacing the other.

What should I look for when evaluating a natural-language sourcing tool?

Index size and freshness (how many profiles, from how many sources, how current), retrieval quality (whether it surfaces the right people, ideally shown by a benchmark), ranking and explainability (whether results are ranked by fit with the reasoning visible), editable criteria (whether you can correct what it infers and re-rank), and where results go next (whether it carries into review, outreach, and your ATS or CRM).