Imported jobs arrive with whatever structure the source had, which is usually not your structure. The auto-tagger fixes that: an ordered list of rules that run on every import and automatically categorize, tag, mark remote, or skip incoming jobs. Rules are deterministic, so the same job always gets the same result, and they always win over the source's defaults. (The AI on your board helps at setup time, when creating an import source from a niche description; the per-job classification itself is done by these transparent rules, so you can always see exactly why a job was tagged.)
Open the rules
In the sidebar, open Distribution and click Import sources.
Click the Rules tab.
On a fresh board the list is empty, with a one-click starter: Add remote normalizer. That seeds an editable rule that marks a job remote when its title says so (remote, anywhere, nationwide, work from home, wfh, home based, virtual, online, distributed, telecommute). It fixes the most common feed gap, do it first if you import remote jobs.
Create a rule
Click Add rule.
Under Condition, give the rule a Name (for example "Categorize engineering jobs").
Pick what to Match against: Job title, Title + description, Company or Location.
Enter Terms, comma- or newline-separated. The rule matches if the field contains any of them, case-insensitive. Leave Terms empty to match every imported job.
Optionally flip Invert the match to match jobs that do NOT contain any of the terms.
Click Test against imported jobs to dry-run the condition against your most recent imported jobs. It lists the matches (title, company, location) without changing anything, so you can tune the terms before saving.
Under Action, pick what happens to matched jobs:
Set category: assigns one of your job categories.
Set job type: assigns one of your job types.
Add tags: comma-separated tags added on top of the job's existing tags.
Mark remote / on-site: sets the remote flag (a switch chooses which way).
Skip the job (don't import): the job is dropped entirely.
Leave Active on and click Create rule.
You get a "Rule created" toast and land on the rule's edit page. Inactive rules are kept but not applied.
How rules run
Rules run top to bottom on every import, and every matching rule applies its action cumulatively: one job can get a category from rule 1, tags from rule 3, and a remote flag from rule 4.
A matching Skip rule stops everything: the job is not imported at all.
Because order matters, drag rules by their handle on the Rules list to reprioritize. Put Skip rules near the top so off-topic jobs are dropped before other rules waste work on them.
Recipe: keep only on-topic jobs. Create one rule with your niche terms in Terms, flip Invert the match on, and set the action to Skip the job. Anything that does not mention your niche never reaches the board.
Re-apply rules to jobs already imported
On the Rules tab, click Re-apply to imported jobs (top right of the list).
All active rules run over every previously imported job, and a toast reports how many were updated.
Use this after adding or reordering rules so your history matches your new setup. Two safety properties: re-applying only touches imported jobs (never ones you posted by hand), and Skip rules never delete existing jobs; skip only prevents future imports.
Rules and the rest of the pipeline
Rules run after the per-source defaults from Import jobs automatically, so a rule's category or type overrides the source default, being more specific. They do not run on jobs you or employers post manually; for those, set classification directly on the job form. Well-tagged imports feed straight into your category and tag landing pages, which is where an aggregated board earns its SEO.