Refolk
ReferenceApplying at volume

The Job Board Search Control Reference, Each Filter Keyed to What It Drops

For any LinkedIn or Indeed search control, name the field it reads, the postings it lets through or drops, and confirm in minutes whether it still works.

19 min readLast reviewed November 21, 2026Read as Markdown

Key takeaways

  • On LinkedIn, f_TPR=r3600 (the "past hour" URL trick) is honoured logged out but, in practitioner tests of the logged-in AI search, returns the same set as the 24-hour preset with no error shown.
  • LinkedIn strips sortBy=DD when the page loads on its new search route, so a "newest first" saved URL quietly falls back to relevance order.
  • A board's "posted X ago" label is the date of the board listing, which is often a refresh or repost, not the date the employer opened the requisition.
  • LinkedIn stores workplace type and location as separate fields, so a job can pass the Remote filter while its location field or description restricts who can apply.
  • Indeed's company: operator can narrow a search to one employer but cannot exclude one, which matters more in the US, where Refolk's index counts 118,143 recruiters, about 15.1x the UK's 7,810.
  • A URL check alone cannot catch a coerced filter; only comparing result counts against a baseline and the nearest preset can.

This is a lookup document for anyone who runs the same saved job searches across several boards every morning and needs to know which controls do what they claim. Each row names a search control (date window, sort order, Boolean operator, workplace type, URL parameter), the field it actually reads, the postings it silently lets through or drops, and a check you can run in a couple of minutes. Jump to the row you need, run the check, and leave.

A note on scope before you start. The evidence below covers LinkedIn and Indeed in depth. For Glassdoor and ZipRecruiter, how the date controls behave has not been publicly established. One useful fact: every job posted to Indeed has also appeared on Glassdoor since the two were linked, so Indeed's behaviour is a reasonable first guess for Glassdoor, but verify it with the procedure further down. The same goes for easy-apply, experience level and job type filters. I found no public test of what they read, so treat them as unverified and run the generic check on them.

What does each job board search control actually read?

Most freshness and scope controls read a board-side field that differs from what the label suggests: the board listing date rather than the requisition date, a workplace tag rather than the job text, a personalised result set rather than a neutral one. The two tables below are the whole reference in compressed form. The sections after them give each row in detail.

Freshness controls

ControlWhat it really readsHow it misleadsQuick check
LinkedIn date posted (f_TPR)Board listing date, in secondsSub-24h values coerced to 24h when logged inr3600 vs r86400 counts, logged in
LinkedIn sort (sortBy=DD)Nothing; stripped on loadList stays in relevance orderReload, read the URL
Indeed date posted (fromage)Board date, in daysDate is often a repostCompare against the ATS
"Posted X ago" labelBoard record dateRefresh or repost looks newOpen the employer's ATS posting

Query and scope controls

ControlWhat it really readsHow it misleadsQuick check
LinkedIn BooleanUppercase AND/OR/NOT, straight quotesLowercase or curly quotes fail silentlyRetype, compare counts
Indeed operatorsDash documented; NOT, title:, company: reportedAdvanced search page goneCount with and without operator
LinkedIn remote (f_WT)Workplace tag, separate from locationTag can disagree with job textSearch body for "hybrid"
Indeed company:One employerCannot exclude an employerNone; no documented exclusion
Indeed duplicate filterNear-duplicate postingsDrops postings silently"Omitted job postings" link
Indeed personalisationYour search history and activitySame search, different countsRun logged out or clear history

Two terms recur throughout. An ATS (applicant tracking system) is the software where the employer actually opens and manages the job; the board posting is usually a copy of it. Coercion means the board accepts a value it does not support, quietly swaps it for one it does, and shows no error.

Date-posted windows and the LinkedIn past-hour URL

LinkedIn's date filter reads the listing date in seconds, and the "past hour" value (f_TPR=r3600) works logged out but, in published tests, is coerced to 24 hours on the logged-in AI search. Indeed's equivalent counts in whole days, so it has no sub-day unit at all.

LinkedIn's interface menu offers only Past 24 hours, Past week and Past month. Behind them sits the f_TPR URL parameter, whose presets are r86400 (24 hours), r604800 (one week) and r2592000 (one month). The number is seconds, which is why the folklore says you can type r3600 and get the past hour.

The folklore was tested in one mode and is used in another. The sources that say "the parameter itself accepts any number of seconds" tested logged out. EarlyApply's logged-in test found that with AI search on, anything under 24 hours is treated as 24 hours, and a second tester (JobLens) reported that on the new search route, r3600 renders the same results as the 24-hour preset. JobLens did confirm that r86400 and r2592000 work on that route. At its global rollout, LinkedIn reported more than 1.3 million members using AI-powered job search daily and over 25 million searches a week, so the mode most people search in is the mode the trick was never tested in.

Board / routeInterface presetsURL parameterSub-24h value honoured?
LinkedIn, logged out24h / week / monthf_TPR=r[seconds]Yes (EarlyApply, sausagedog)
LinkedIn, logged in (AI)24h / week / monthf_TPRNo, coerced to 24h (EarlyApply; JobLens)
LinkedIn sortRemovedsortBy=DD strippedn/a
Indeed24h / 3d / 7d / 14d (older doc)fromage=[days]No sub-day unit; untested publicly

On Indeed, the presets in older documentation were Last 24 hours, Last 3 days, Last 7 days and Last 14 days, and the fromage parameter counts days (fromage=1 returns the last 24 hours). Nobody has published a test of values below one day. Since the unit is days, I would not expect one to work, but that is an inference, not a finding.

There is also a live source conflict. One URL-builder tool claims the Date posted filter can look empty in the interface while the results still follow the window in the URL. That contradicts the logged-in tests above. I cannot tell you which is right for your account today, and neither can anyone else without testing. That is why the verification procedure below compares counts rather than trusting what the interface or the URL shows.

Which date the "posted X ago" label shows

The posted label and every date filter read the board's own listing date, which is often a refresh or repost, not the date the employer opened the role. A "posted 2 hours ago" job can be weeks old at the source.

LinkedIn's "posted X ago" is often a refreshed or repost date, not the original. Indeed has stopped putting the date on its job pages, and the date it uses is often a repost. The mechanism behind this is in Greenhouse's support documentation: LinkedIn and Indeed limit organisations from refreshing or "bumping" a free job post through Greenhouse Recruiting. So when an employer wants a stale free listing back near the top, the practical route is to resubmit it, which LinkedIn marks "reposted", and the date filter then reads the resubmission date as the job's age.

What it proves: the listing was created or resubmitted on the board inside your window. What it does not prove: that the role is new, or that nobody has been interviewed yet. How far board dates typically drift from ATS dates has not been publicly established. Measure it on your own target employers with step 6 of the procedure.

How to sort jobs by most recent

On LinkedIn's current search route there is no documented way to sort by newest: the sortBy=DD parameter is stripped when the page loads, and the classic page that honoured it has been removed. Use a tight date window run on a fixed schedule instead of a sort.

JobLens reported that the new interface strips sortBy, and that the classic search page that honoured sortBy=DD is gone. Refolk's own blog tracked the classic job search beginning its retirement on the member side. The failure state is easy to spot if you look: reload a saved URL that contains sortBy=DD and read the address bar after the results render. If the parameter is missing, your list is in relevance order no matter what your bookmark says.

The substitute that holds up is a narrow window you can verify, the 24-hour preset, combined with a fixed run time each morning. A 24-hour window run daily covers every listing once without depending on a sort that no longer exists.

Boolean operators on LinkedIn and Indeed

LinkedIn documents uppercase AND, OR and NOT, straight quotes and parentheses, and nothing else is officially supported. On Indeed, the advanced search page is gone and the replacement is operators typed into the search bar, with documentation that disagrees with itself.

LinkedIn

LinkedIn's help page "Use Boolean search on LinkedIn" is the authority here. Its rules:

  • AND, NOT and OR must be typed in uppercase. Lowercase and, or, not are treated as ordinary words.
  • Quotes and parentheses are supported.
  • Braces { }, brackets [ ], angle brackets < > and wildcards such as * are not supported.
  • The + and - operators "may appear to work in some cases" but are not officially supported.
  • title: and company: are not on LinkedIn's help pages today.
  • Free accounts have an operator cap, but the help page does not give a number.

Curly quotes are the quiet killer. Paste a query from a word processor and the straight quotes become curly ones, which break phrase searches without any error. LinkedIn has also not documented which operators its AI search mode ignores; that is not publicly established, so test any query you depend on in the mode you actually use. (In LinkedIn Recruiter, the Project filter does not support Boolean search at all, which matters only if you also search from the recruiter side.)

Indeed

Indeed's documentation disagrees about exclusion: its career guide lists NOT, but its help centre only documents the dash. Practitioners report that title: and company: work. One hard limit is documented by practitioners: company: narrows results to one employer but cannot remove one. There is no documented way to exclude a single employer from an Indeed search.

Use this query skeleton as a starting point. It sticks to the documented syntax on each board.

Board-safe query skeletons
LinkedIn (documented syntax only):
("data engineer" OR "analytics engineer") AND (Python OR SQL) NOT (intern OR internship)

Indeed (dash for exclusion, as the help centre documents):

Indeed (single employer only; company: cannot exclude):

Swap in your own titles and skills; keep the capitals and straight quotes exactly as shown.

Workplace type, location, and why the remote filter shows hybrid jobs

LinkedIn stores workplace type and location as two separate fields, so the Remote filter checks the tag, not where you are allowed to live. A job can pass the filter correctly and still exclude you through its location field or its text.

Every job posted on LinkedIn must have a valid location. For remote jobs, the employer can choose a city, metro, state or country, and LinkedIn's location field also accepts a country cluster such as the European Union. Once a job is posted, its location cannot be edited. So a "Remote" job located in one state, or in the EU cluster, is remote within that area, and the filter says nothing about that restriction.

Jobs imported from an employer's ATS through LinkedIn's Job Wrapping get their workplace tag one of two ways: from #LI tags the employer adds, or from an auto-tagger whose inputs include the job title text, job description text and job location text. The auto-tagger is on by default. Either source can disagree with the description, which is where the remote-tagged job that says "hybrid, three days in office" comes from.

Where a remote job's real restriction lives

  1. Workplace tag
    Remote, hybrid or on-site; set by #LI tags or a default-on auto-tagger
  2. Location field
    City, state, country or a cluster such as the European Union; cannot be edited after posting
  3. Job description text
    Phrases like hybrid, office, must reside that override the tag in practice
  4. Employer's ATS posting
    The source record the board copied from
The Remote filter reads only the top layer, while the restrictions that decide your eligibility sit in the layers underneath.

The check is mechanical. Before applying to anything the Remote filter surfaced, search the posting body for "hybrid", "office" and "must reside", and read the location field. The share of remote-tagged postings that carry a location restriction has not been publicly established, so do not assume it is rare.

Do not try to infer remote-friendliness from the people side either. In Refolk's index, only 180 of 118,143 US recruiters put "remote" in their headline.

0.15%
US recruiters in Refolk's index with "remote" in their headline
180 of 118,143 people holding Recruiter, Technical Recruiter or Talent Acquisition titles, so a remote signal has to come from the posting, not the person.

Indeed's hidden drops: duplicates, personalisation, and agency postings

Indeed quietly removes near-duplicate postings, personalises results to your history, and offers no documented way to exclude an employer. Together these mean two people running the same saved search can see different lists, and neither sees everything.

Duplicate collapse

Indeed collapses near-duplicates and says so only at the end of the list: "We have removed some job postings very similar to those already shown." Its legacy API had a duplicate filter that was on by default, where 0 turned it off. On the website, the check is the "omitted job postings" link at the bottom of the results. Click it whenever the same role is posted by several agencies, because the version Indeed kept may not be the one you want.

Personalisation

Indeed's help centre states: "Search results are customized based on your previous searches and your activity on Indeed." A saved search is therefore not a fixed query; its counts drift as your history changes. Indeed documents clearing search history as a step, and running the search logged out gives you a neutral comparison.

Indeed's "Not interested" preferences are unsettled. One help page says preferences filter out matching jobs, and another says they will not change search results. Test it: mark one title Not interested, re-run the search, and compare counts.

Indeed has also stopped sending daily job feed emails, so if a morning email was your freshness check, it is no longer arriving and the saved search is now the whole system.

Agency postings and the missing exclusion

The company: limit bites hardest in the US. In Refolk's index, the US recruiter population is about 15 times the UK's, and in small samples the most common US employers were staffing firms while the UK's were in-house teams.

MarketRecruiters in Refolk's index"Technical Recruiter" in sample of 25Generic "Recruiter" in sample of 25
United States118,14310 (40%)8 (32%)
United Kingdom7,8104 (16%)13 (52%)
US:UK multiple15.1x--

In the 25-person US sample, the most common employers were the staffing firms K2 Partnering Solutions (3) and Experis (2). In the UK sample they were in-house teams at IKEA (2), Utility Warehouse (2) and Oceaneering (2). Twenty-five people is a direction, not a finding, but it points the same way as the tooling gap: in a US search you will see agency reposts, Indeed may collapse some of them, and company: cannot strip the rest.

15.1x
US recruiter pool relative to the UK in Refolk's index
118,143 US versus 7,810 UK people with Recruiter, Technical Recruiter or Talent Acquisition titles; the sample percentages beside it are indicative only.

When the same role arrives from three agencies, the useful question is no longer which filter to trust but which version to answer. Refolk scores how well you fit each posting and tailors your resume to it, so you can compare near-identical listings on their actual requirements instead of their posting date.

How these controls go wrong: failure modes and false positives

Every failure below is silent: the board shows a normal-looking list, no error, and often the parameter you typed still in the URL. That is why the checks compare counts and source dates rather than reading the interface.

A filter that fails loudly costs you a minute. A filter that fails quietly costs you a week of stale applications.
  1. f_TPR=r3600 while logged in. Returns the 24-hour set with no error, so the "past hour" list is a false positive. Check: run r3600 and r86400 logged in. Identical counts mean the value was coerced.
  2. sortBy=DD. Stripped on load; the list stays in relevance order. Check: reload and read the URL.
  3. "Posted X ago". The board's record date, which may be a refresh or a repost. Check: open the employer's ATS posting and compare dates.
  4. Remote filter. The tag comes from a default-on auto-tagger or from #LI tags, and either can disagree with the description. False positive: a remote-tagged job whose text says "hybrid" or names a city. Check: search the body for "hybrid", "office" and "must reside".
  5. Lowercase operators or curly quotes on LinkedIn. Lowercase and/or/not become ordinary words; curly quotes wreck phrase searches without any error. Check: retype with uppercase operators and straight quotes and compare counts.
  6. Indeed near-duplicate collapse. Postings dropped silently. Check: the "omitted job postings" link at the end of the results.
  7. Indeed personalisation. The same saved search returns different counts as your history changes. Check: clear search history or run logged out.
  8. Indeed "Not interested". Help pages contradict each other on whether it filters results. Check: mark a title, re-run, compare.

The matrix below turns the two readings you take in the procedure, whether the parameter survived in the URL and whether the count moved, into a verdict.

Reading a filter check

Count changes as expectedCount identical to preset or baseline
Rewritten
Rebuild the saved search from the interface menu and recheck
Working
Log a pass and recheck next week
Removed
Drop the control from saved searches, as with sortBy=DD
Silently coerced
Treat it as the nearest preset, as with r3600 logged in
Parameter stripped from URLParameter kept in URL
A parameter that stays in the URL while the count matches the nearest preset is the most dangerous state, because everything looks correct.

No board publishes a verification procedure, and nobody has published a result-count threshold that means "filter ignored". Identical counts between two values that should differ is the cleanest signal available. A small difference is harder to read, because personalisation and new postings arriving between runs also move counts, so run the two searches back to back.

Run a baseline, apply the filter through the menu, edit the value, read the URL after load, then compare counts across logged-in and logged-out sessions and spot-check dates against the employer's ATS. Steps 2, 3 and 6 are documented by practitioners; the rest, and the time estimates, are my own.

The filter verification pass

  1. Run a baseline
    Run the saved search with every filter off and write down the result count. About one minute.
  2. Apply the filter in the interface
    Set the filter through the board's menu and confirm its parameter (f_TPR, fromage, f_WT) appears in the URL. Tutorials document this order: interface first, then edit the URL.
  3. Edit the value in the URL
    Change the value you want, for example r86400 to r3600, and reload. Done when the edited value is still in place after reload.
  4. Read the URL after the page loads
    Check for parameters the board removed, such as sortBy. Done when each one is confirmed present or stripped.
  5. Compare logged out and logged in
    Run the same URL in a private window and against the nearest preset. Matching counts mean the value was coerced; differing counts mean it was honoured. About two minutes.
  6. Check posted dates against the ATS
    Open a sample of postings on the employer's careers page or ATS and log the gap between the board date and the source date. I suggest 20; that number is mine, not documented.
  7. Re-verify weekly
    Repeat for every control your saved searches depend on. Done when each control has a dated pass or fail line.

EarlyApply puts the logged-out comparison first; most other guides skip it entirely. I put it after the URL checks because you need the edited URL before there is anything to compare, but the order matters less than doing it at all. It is the only step that catches the r3600 failure.

Keeping this reference current

Boards change these controls without announcements, so the only reliable version of this table is the one you re-verify on a schedule. Keep a dated log, one line per control, and rerun the pass every week.

The evidence for that habit is in this document. Indeed's advanced search page started returning a not-found message with no announcement I could find. LinkedIn's sortBy stopped working when the classic route was retired. The r3600 trick works or fails depending on which mode you are in. None of these changes produced an error message. Paste this log into whatever you track your search in:

Weekly search control log
Date | Board | Control | Value tested | Logged-in count | Logged-out count | Preset count | Param in URL after load? | Verdict (PASS / COERCED / STRIPPED) | Notes
____ | LinkedIn | f_TPR | r3600 | ____ | ____ | r86400: ____ | Y / N | ____ | ____
____ | LinkedIn | sortBy | DD | ____ | ____ | n/a | Y / N | ____ | ____
____ | LinkedIn | f_WT | remote | ____ | ____ | n/a | Y / N | ____ | hybrid/office/must reside hits: ____
____ | Indeed | fromage | 1 | ____ | ____ | n/a | Y / N | ____ | omitted postings: ____
____ | Indeed | company: | ____ | ____ | ____ | n/a | Y / N | ____ | ____

One line per control per week; a FAIL line means you rebuild that saved search before trusting it again.

Once the list is trustworthy, the time you save on filtering should go into the applications themselves. Refolk drafts a tailored resume and cover letter for each posting you keep and scores your fit, so a cleaner morning list turns into more sends rather than more scrolling.

Before you trust a saved search

  • Every saved search has a baseline count recorded with all filters off.
  • Any sub-24-hour f_TPR value has been compared against r86400 while logged in.
  • No saved URL relies on sortBy=DD surviving page load.
  • Every LinkedIn query uses uppercase AND, OR, NOT and hand-typed straight quotes.
  • Remote-tagged postings are checked for hybrid, office and must reside before applying.
  • The Indeed omitted job postings link has been opened at least once for each saved search.
  • At least one Indeed search has been run logged out to measure personalisation drift.
  • A sample of posted dates has been compared against the employers' own ATS postings.
  • Each control has a dated PASS, COERCED or STRIPPED line from the past seven days.

Questions job seekers ask

Does the LinkedIn past hour URL (f_TPR=r3600) still work?

It works logged out and, in practitioner tests, fails logged in on the AI-powered search. The f_TPR parameter counts seconds, so r3600 means one hour, and logged out the board honours any number of seconds. Logged in, testers found anything under 24 hours is treated as 24 hours, with r3600 rendering the same results as the 24-hour preset. Check your own session by comparing r3600 against r86400: identical counts mean it was coerced.

How do I sort LinkedIn jobs by most recent now?

There is no documented way to force a newest-first sort on LinkedIn's new search route. The old sortBy=DD parameter is stripped when the page loads, and the classic search page that honoured it has been removed. The closest substitute is a tight date window, such as the 24-hour preset, run every morning. Reload any saved URL and read the address bar to confirm whether sortBy survived.

Where did Indeed advanced search go?

Indeed's advanced search page now returns a not-found message, and I could find no announcement from Indeed about removing it. The replacement is typing operators straight into the search bar. Indeed's help centre documents the dash for exclusion, its career guide also lists NOT, and practitioners report title: and company: work. Test each one on your own search by comparing counts with and without the operator.

Why does the remote filter show hybrid jobs?

Because the remote tag and the job text are set separately, and they can disagree. On LinkedIn, jobs imported from an employer's applicant tracking system get their workplace tag from #LI tags or from an auto-tagger that reads the title, description and location text and is on by default. A posting can carry the Remote tag while its body says hybrid or names an office. Search the description for hybrid, office and must reside before applying.

Why are my Boolean operators not working in job search?

On LinkedIn the usual causes are lowercase operators and curly quotes. AND, OR and NOT must be uppercase or they are treated as ordinary words, and curly quotes pasted from a document break phrase searches without any error. LinkedIn also does not support wildcards or brackets, and + and - are not officially supported. Retype the query with uppercase operators and straight quotes, then compare the counts.

Put this to work

Paste your career in once. Every application after that is written for you.

Drop a resume or a LinkedIn URL. I rank the live openings against it, rewrite the resume and write a cover letter for the best of them, and fill in the employer's form when you press the button. You read, you decide what goes out.

  1. 01Drop your resume

    A PDF or a LinkedIn URL. About a minute, once.

  2. 02I rank the openings

    Every weekday morning, the live catalog scored against your history. Up to 20 worth your time, not two hundred links.

  3. 03Each one is written up

    Resume rewritten for the posting, a cover letter, a fit score. Press send, or let me fill in the form.

  • New matches ranked and written before you are up.
  • Every bullet stays inside what your history supports. Nothing invented.
  • Queued, submitted, interviewing, offer: one screen, not a spreadsheet.

Free, no card. Up to 25 applications sent a week, and nothing is sent until you say so.

Read next