Build Automated Job Search Alerts With Career Page Monitoring
Build automated job search alerts that monitor career pages, filter the right roles, remove duplicate matches, and email only genuinely new opportunities.
TABLE OF CONTENTS
Automated job search alerts work best as a pipeline, not as a single scraper. Use an Apify Actor to collect jobs from each target career page, normalize the results, filter them with clear role logic, compare each match with a Google Drive spreadsheet, and send an email only when the job has not appeared before. This approach keeps persistent listings from generating repeat alerts and helps prevent overlap between company career pages and broader job databases.
The workflow at a glance
A practical system has seven stages:
- Collect: Run an Actor for each career page or use an Actor that accepts multiple sites.
- Normalize: Map different site formats into consistent fields such as company, title, location, description, and job URL.
- Filter: Apply keyword rules to titles and descriptions before anything reaches the alert step.
- Match: Build a stable key for each opening, starting with the canonical job URL.
- Deduplicate: Check the key against a shared spreadsheet of previously seen jobs.
- Store: Append only new matches to the spreadsheet.
- Notify: Email the new rows, not the full scrape result.
Apify describes Actors as programs that take structured input and can run manually, through an API, or on a schedule. Its datasets are designed to hold structured scraping results, which makes the platform a useful collection layer for this workflow.
Build automated job search alerts step by step
1. Start with a small test group
Begin with four career pages that represent the sites you care about most. Don't try to monitor every possible employer on day one. A small test group makes it easier to identify which pages return clean job records, which require extra instructions, and which need a site-specific scraper adjustment.
For every first run, verify:
- the Actor reaches the intended career page;
- job titles and URLs are captured;
- locations and descriptions are not being dropped;
- closed or unrelated roles are excluded when possible; and
- the output is consistent enough to compare with later runs.
Keep a short failure log. A page that works today may change its markup later, so the log becomes your starting point when a scheduled run suddenly returns zero jobs.
2. Define keyword rules before you scrape at scale
Broad keywords create noisy alerts. Use a small set of high-signal terms, plus compound rules for phrases that only matter in context.
For a corporate real estate search, a rule set might look like this:
- include roles containing occupier;
- include real estate only when it appears with lease administration or portfolio strategy;
- exclude unrelated roles that mention property only as a benefit, location, or generic company term; and
- keep title and description matching separate so you can tune them independently.
Write the rules as explicit conditions rather than one long search string. That makes it easier to see why a role matched and to adjust one condition without breaking the rest of the workflow.
3. Configure and test each Actor
Give each Actor the smallest useful input: a career-page URL, the fields to extract, and the filtering settings you want to test. Run it manually first. Inspect several results, including a likely match and a likely non-match, before adding a schedule.
Career pages rarely share one structure. Some use common applicant-tracking systems, while others require custom navigation or a different field selector. If one site fails, adjust that site’s instructions instead of weakening the rules for every other source.
Once the test run is producing reliable output, schedule the Actor daily or weekly based on how quickly openings change. Apify recommends running an Actor at least once before scheduling it, and schedules can use cron expressions and a selected time zone.
4. Choose a deduplication key
The deduplication key determines whether a job is treated as new. Use this order of preference:
- canonical job URL;
- source URL plus normalized title;
- company, normalized title, and location as a fallback.
Normalize URLs by removing tracking parameters when safe. Normalize titles by trimming whitespace and standardizing case. Keep the original title and URL too, because the original values are what a reader needs in the email.
A spreadsheet is enough for an initial system. Useful columns include:
- first seen date;
- last seen date;
- source company;
- job title;
- location;
- canonical job URL;
- deduplication key; and
- alert sent date.
Google Sheets includes a built-in Remove duplicates cleanup action, but an automated workflow should still check the key before sending an alert. Manual cleanup is a useful repair tool, not a substitute for a pre-alert comparison.
5. Separate collection from notification
Don't email every row a scrape returns. First compare the filtered results with the spreadsheet. If the key already exists, update the last-seen date if needed. If it does not exist, append the row and pass only that new record to the email step.
This two-stage design prevents a job that remains open for several weeks from producing a new message every day. It also gives you one place to compare results when a broad job database and direct company-page monitors find the same opening.
6. Scale from four pages to a larger list
After the first group is stable, add sites in batches. A repeatable pattern is more valuable than a large first launch:
- duplicate the working Actor configuration;
- replace the source URL;
- run the new Actor manually;
- inspect the output and failure logs; and
- add the site to the scheduled group only after it passes.
For a list of roughly 40 sites, expect a mix of clean first-pass results and pages that need individual adjustments. Review the first batch promptly so broken selectors do not go unnoticed for days. Keep the same normalized output and deduplication rules across every source.
Troubleshooting common problems
The Actor returns no jobs
Confirm that the career-page URL is public and still leads to the job listings. Check whether the site requires a second click, JavaScript rendering, or a different page path. Compare the current page with the URL used in the last successful run.
Too many unrelated roles match
Tighten the rule set. Replace broad terms with title-and-description conditions, add required companion phrases, and maintain a small exclusion list for common false positives.
The same job arrives twice
Inspect the URLs and titles from both sources. If one source adds tracking parameters or a company name to the title, normalize those fields before comparing keys. When the source does not provide a stable URL, use a composite key and keep the original fields for review.
A site breaks after working
Treat each career page as its own maintenance unit. Review the failure log, test the Actor manually, and update only the affected configuration. Don't change the shared filter or notification logic unless the problem is global.
A maintenance checklist
Review the workflow monthly, or sooner if hiring volume is high:
- confirm each scheduled run completed;
- check for a sudden drop to zero results;
- review new false positives and false negatives;
- remove stale tracking parameters from keys;
- verify the spreadsheet is still receiving new rows; and
- test that the email step sends only unseen matches.
The strongest setup is the one you can inspect and adjust without rebuilding the whole system. Keep collection, filtering, deduplication, and notification as separate steps, then expand the source list only after the first group is dependable.
Get help building this
Job alert pipelines usually break at the site-specific layer: a career page changes its markup, a scheduled run quietly returns zero jobs, or two sources report the same opening under slightly different titles. If you've hit that wall, book a ZoomFlow session. One of our consultants can work through your source list, filter rules, and deduplication logic with you live.