AI Features for Candidates

Pro feature surface. These flows activate when WP Career Board Pro is active and the relevant AI provider is configured. On Free the keyword search and standard apply flow still work.

This page covers what candidates experience with AI on (1.4.3): the AI Chat Search, AI candidate matches (“Recommended for you”), and the AI cover-letter writer in the apply panel.

AI Chat Search

A chat-style natural-language search box that returns semantically relevant jobs - not just keyword matches.

Where it lives

The AI Chat Search block (wcb/ai-chat-search) is placed on a page via the block editor. When Pro sets up, an AI Job Search page carrying this block is created automatically. The site owner can also add the block to the Find Jobs page, above the standard keyword search. If your board doesn’t have it, ask the admin to add the block.

What candidates type

Anything natural-language. Examples that work well:

“Remote React job, US time zones, $100k+”
“Junior data analyst position, willing to relocate to Berlin”
“Marketing role at a B2B SaaS startup, 10-50 people, hybrid OK”
“Senior backend engineer, Rust or Go, no on-call”

What doesn’t work as well:

One-word queries (“developer”). For a single word, the standard keyword search is faster and more predictable.
Queries that depend on context outside the listings (“the job from last week’s newsletter”). The AI doesn’t know about your other pages.

What candidates see

The block presents a conversation-style box:

The typed query appears as a message bubble.
A spinner for 1-3 seconds while the query is embedded and matched.
An assistant reply noting how many jobs matched, followed by a list of matched job cards (title, company), ranked by relevance, not date.

The results render in-place via the Interactivity API - no page reload happens when they refine the query. If a search fails, the box shows a “Search failed. Please try again.” message rather than breaking.

What it can’t do

It won’t enforce filters you didn’t type. “Any job” returns the top jobs by general relevance. The standard keyword filters still work as a fallback.
It doesn’t store search history per candidate. Each query is embedded, matched, and forgotten on the backend.
It doesn’t auto-recommend jobs by email. Job Alerts use keyword / filter match, not AI semantic match.
It needs job embeddings to exist. Embeddings are generated at wcb_job_created. Jobs from before AI was enabled appear only after the admin runs Index existing jobs (see 02-setup-and-providers.md).
It has a per-user rate limit: 30 AI calls per hour across all AI features. Heavy testers will hit a 429 and need to wait.

Tips for candidates

Be specific about non-negotiables. “Remote” surfaces; “no on-call” surfaces; “must use modern frameworks” is too vague.
Salary specifics help. “$80-100k” filters better than “good salary.”
Include location AND remote preference. “Remote, US time zones, preferably East Coast” is a complete instruction.
If results feel off, refine - add one more constraint and resubmit.

AI candidate matches (“Recommended for you”)

Beyond typing a query, AI can match jobs to a candidate’s resume automatically. The candidate’s resume (built in the Pro Resume Builder) is flattened into text, embedded, and compared against every job vector. The top jobs are returned as “Recommended for you” cards.

This runs through GET /wcb/v1/candidates/{id}/matches for a specific candidate, and POST /wcb/v1/ai/match for the currently signed-in candidate. It needs an embedding provider (OpenAI or Ollama) configured and job embeddings to exist.

Quality depends on how complete the candidate’s resume is - the more sections filled in (experience, skills, summary), the sharper the matches.

AI cover-letter writer

In the job apply panel, a candidate can generate a tailored cover letter instead of writing one from scratch.

The flow

Open a job and start to apply.
Click Generate with AI next to the cover-letter field.
Pro sends the candidate’s resume text plus this job (title and a content excerpt) to the analysis provider via POST /wcb/v1/jobs/{job_id}/ai-cover-letter.
A first-person draft (about 150-200 words, using only resume-supported details) is inserted into the cover-letter field.
The candidate edits the draft, then submits the application as normal.

Notes

The button appears only when Pro is active and the analysis provider is configured (the apply panel reads an aiCoverEnabled flag).
The draft uses only details the resume supports - it won’t invent experience. A sparse resume produces a thin letter.
Same 30-calls-per-hour rate limit applies.
If the candidate applies with a Resume Builder resume that has no PDF, Pro generates and attaches the PDF automatically, so the application still carries a downloadable file.

How resume data reaches the AI

Pro’s Resume module implements the wcbp_candidate_resume_data filter: it returns the candidate’s grouped resume sections, which the AI module flattens into text for matching, ranking, and cover letters. This is wired in Pro core - no add-on is required. Resumes imported from WP Job Manager Resumes are read too (legacy experience / education meta), so matching and cover letters work on migrated data.

There is no separate “auto-parse an uploaded PDF into profile fields” step - candidates fill in the Resume Builder (or import), and that structured data is what feeds the AI.

What candidates get without Pro

Capability	Free	Pro 1.4.3
Browse all jobs	Yes	Yes
Keyword search + filters	Yes	Yes
Apply to jobs	Yes	Yes
Resume upload (attaches to applications)	Yes	Yes
Save jobs to bookmarks	Yes	Yes
Get email notifications	Yes	Yes
Natural-language AI search (AI Chat Search block)	No	Yes
“Recommended for you” AI candidate matches	No	Yes
AI cover-letter writer in the apply panel	No	Yes
Semantic match in job alerts	No	No

Where to go next

04-employer-ai-features.md - the other side of the table.
02-setup-and-providers.md - if you’re also the site admin.