Importing a job from a URL

Available to beta program members, plus Military, Student, and Admin accounts. Turn the beta on in Settings.

Paste a job posting URL and we fill in the job title, company, and description for you, which saves you switching tabs and copying three fields across by hand. It runs while you wait.

Image coming soon

Where to find it

It sits at the top of the Job Details column in the Smart Scan flow, under the line "Paste a job URL and we'll fill this in automatically."

It is part of the Smart Scan flow, so start a Smart Scan when you want to import from a URL.

How to use it

1. Paste the posting URL into the field. The placeholder shows the shape we expect, for example https://www.amazon.jobs/en/jobs/2730539.
2. Press Extract. Once it has run, the button reads Extracted.
3. Read the summary line that appears underneath, which tells you what we found and what we did not: a job title, a company, and a description of a given length.
4. Read the job description field before you scan. Postings vary in how they are built, and a quick look tells you whether we pulled the requirements or stopped short of them.

Which sites work

Over 100 job boards and applicant tracking systems have their own configuration, including LinkedIn, Indeed, Glassdoor, ZipRecruiter, Monster, Dice, SimplyHired, CareerBuilder, Seek, and USA Jobs, plus the systems companies host their own postings on: Greenhouse, Lever, Workday, Taleo, iCIMS, SmartRecruiters, Jobvite, JazzHR, BambooHR, BreezyHR, Ashby, ADP, Paycom, Paylocity, Paycor, Dayforce, SuccessFactors, Oracle HCM Cloud, Cornerstone, and Avature.

Anything not on that list falls back to a generic reader that looks for the standard job posting markup most career sites publish, which is why unlisted sites often still work.

A few URLs are refused before we try, because they cannot be job postings in the first place: links to files (PDF, DOC, DOCX, XLS, XLSX, ZIP) and Google Docs links. Paste the URL of the posting page itself instead.

When it does not work

Some sites block automated readers and some require a login before they will show a posting, and neither is something we can read around. When that happens you will see one of these:

• "site is a private job board that our parser cannot access."
• "site requires a login to view this posting."
• "site is blocking our automated parser."
• "site has a security certificate problem, so we cannot reach it safely."
• "We could not read the full job description from site."
• "We could not find a job listing at that URL. Please check the link is correct and still open."

In every case the fix is the same and it is quick: copy the job description from the posting and paste it below. The scan reads a pasted description exactly the same way it reads an extracted one, so nothing about your report changes.

Tell us when a parse is wrong

Reporting a bad parse is how a specific site gets fixed, and it takes about 20 seconds.

• If the extract failed for a reason that is not the site's own doing, a report button appears next to the error.
• If it worked but the result looks wrong, use Doesn't look right? under the summary.

Either one opens Report this job URL, which explains what it sends: "We'll file a bug report with the URL and everything we pulled from it, so we can fix this site." Fill in What went wrong?, for example "the company name is wrong, or the duties section is missing", and press Send report. You will get back "Thanks, we've logged this URL and we'll get the parser fixed."

Reports go straight to our bug tracker with the URL attached, which is what lets us fix that specific site rather than guessing.

How the extraction actually works

The order of operations is worth knowing, because it explains which message you get when a site does not cooperate.

1. We fetch the page normally first. For a small number of sites that never works, so we skip straight to step 2.
2. If that returns nothing usable, we open the page in a headless browser and try again. "Usable" means we found a title and a description of at least 100 words.
3. Each attempt tries the site's own configuration first, then standard job posting markup, then a general read of the main content block.
4. If the result is still missing a title or the description is still under 100 words, and we have some page text to work with, we hand it to AI to pull the description out.

AI is the last resort rather than a step everything goes through, which is also why a blocked page never reaches it. A block means we came away with no page text at all, so there is nothing for the AI to read and you get the block message instead.