WFA Research
How often remote job listings state what they pay, measured from both the salary field and the posting text.
Updated September 3, 2026
Whether a job advert says what it pays is easy to ask and surprisingly hard to measure, because structured salary fields do not capture every disclosure.
An applicant tracking system can expose a structured salary field, and that field is simple to read. But an employer can also write "Salary Range: $120,000 — $150,000" in the posting itself and never touch the field. Counting only the field misses those entirely.
A listing counts here as disclosing pay when either is true: the applicant tracking system exposes a salary field, or the posting text states a pay figure beside an unambiguous pay label. Both are things the employer published.
Under that definition, 44.9% of 19,563 active remote listings from 2,872 employers state pay. Counting only the structured field would have put it at 22.7%.
Two routes, one of which structured field analysis misses.
The two routes are exclusive as counted here: the text figure counts only listings with no structured field, so they add to the total disclosing.
The 4,348 listings in the middle row are the reason this report exists. They state a pay range in plain sight, and a measurement that reads only the applicant tracking system records them as silent.
Among listings whose employer stated a geographic hiring scope, those naming the United States disclose pay 3.0 times as often as those naming other countries but not the United States.
6,189 listings name the United States; 2,477 name at least one country but not the United States. Listings with no stated scope are excluded from both groups.
Several US states and local jurisdictions require a pay range in job postings, and that is the obvious explanation to reach for. The data show an association between US hiring scope and higher disclosure; they do not isolate the effect of legislation from employer mix, occupation, company size or anything else, and this analysis does not attempt to.
Pay transparency legislation has its own page in this series. Here it is context, not a finding.
The same four categories as the accessibility report, and the same snapshot, so the two studies can be read side by side.
Categories below 100 listings are not shown.
| Field | Listings | Stating pay | Disclosure rate |
|---|---|---|---|
| Healthcare | 1,046 | 584 | 55.8% |
| Legal | 402 | 217 | 54.0% |
| Data Science | 910 | 441 | 48.5% |
| Security | 243 | 117 | 48.1% |
| other | 1,699 | 809 | 47.6% |
| consulting | 423 | 200 | 47.3% |
| Project Mgmt | 490 | 226 | 46.1% |
| AI & ML | 419 | 191 | 45.6% |
| Operations | 1,043 | 471 | 45.2% |
| Sales | 2,690 | 1,206 | 44.8% |
| Development | 5,117 | 2,277 | 44.5% |
| Education | 132 | 58 | 43.9% |
| Finance | 670 | 292 | 43.6% |
| Customer Success | 852 | 366 | 43.0% |
| HR | 284 | 121 | 42.6% |
| Product | 777 | 330 | 42.5% |
| Marketing | 968 | 411 | 42.5% |
| Design | 387 | 147 | 38.0% |
| Support | 446 | 166 | 37.2% |
| Content | 248 | 79 | 31.9% |
| Admin & VA | 143 | 40 | 28.0% |
| coaching | 144 | 24 | 16.7% |
The field table covers 19,533 of the 19,563 listings. Withheld for having fewer than 100 listings: ecommerce (30 listings). A listing belongs to one field, so these groups do not overlap.
The obvious next question is what the disclosed jobs pay. We are not answering it with a single number, and the reason is worth showing rather than asserting.
A pay period is recorded on only 481 of the 4,433 priced records as annual, with 3,521 stating no period at all. Treating an unstated amount above $20,000 as annual is an inference, and it moves the answer a long way.
| Definition | Priced listings | Median advertised floor |
|---|---|---|
| Only listings that state an annual period | 473 | $104,500 |
| Adding those inferred annual above $20,000 | 3,832 | $140,000 |
The two differ by $35,500. That is too much to put a single figure in a headline.
| Treat unstated amounts above | Priced listings | Median advertised floor |
|---|---|---|
| $15,000 | 3,838 | $140,000 |
| $20,000 | 3,832 | $140,000 |
| $25,000 | 3,826 | $140,000 |
| $30,000 | 3,825 | $140,000 |
Moving the threshold between $15,000 and $30,000 changes almost nothing. What changes the answer is whether inferred-period records are admitted at all, which is a decision about the data rather than a parameter to tune.
Hourly rates are never converted to annual. Doing so would require assuming hours per week and weeks per year, and neither is in the posting.
| Stated pay period | Priced listings |
|---|---|
| Annual | 481 |
| Hourly | 286 |
| Monthly | 144 |
| No period stated | 3,521 |
Base pay is also kept apart from on-target earnings, commission and bonus. Where a median is computed at all, it uses base pay only.
Every share above counts listings. An employer with many open roles contributes more observations than one with a single role, so the same question asked per employer gives a different answer.
2,872 employers. The ten largest hold 6.9% of the listings and the twenty-five largest 13.1%.
These are three different questions and the middle one is the most generous reading available. None of them is a substitute for the others.
It can tell you how often listings state pay under the definition above, how that varies by hiring scope and by field, and what ranges employers advertised in this sample.
A listing counts as disclosing pay when the employer's applicant tracking system exposes a salary field, or when the posting text states a pay figure beside an unambiguous pay label. Both are things the employer published.
This single definition is used by the page, the charts, the CSV and the structured data. No figure here counts only the structured field.
Snapshot September 3, 2026, methodology version 2.0: 19,563 active listings from 2,872 employers. Listings come directly from employers' own applicant tracking systems, never from an aggregator. That collection favours employers on a modern applicant tracking system, which skews towards technology and English-language postings. Read every figure as describing the remote job market WFA Digital observes.
A money figure counts only when an unambiguous pay label sits beside it within about 110 characters — "salary range", "base pay", "hiring range", "this position pays", and equivalents in a few other languages. A figure with only a weak signal nearby, a bare "compensation" or "per year", is deliberately not counted.
That tier was tested and rejected: reviewing its matches found it mixing genuine pay with education allowances, internet reimbursements and learning budgets, at roughly 70% precision. Under-counting is the acceptable error; inventing disclosure is not.
The detector was developed against 405 listings read by hand, in two rounds. Those rounds found and fixed gaps, so they cannot also measure the result. Everything below comes from a holdout drawn on 2026-09-03, after the patterns were frozen, excluding every listing read during development and allocated across applicant tracking systems in proportion to the pool.
Manual validation of 180 final detections found 98.9% precision (178 of 180, 95% CI 96.0% to 99.7%). Each was read against its quoted evidence, sampled from the 4,254 detections in the holdout pool. Two were wrong: one matched the word "remuneration" beside an education allowance, with "annual salary" appearing only as a life-assurance multiplier; the other carried a "Base Pay" heading whose rate was explicitly left "to be finalized upon selection", quoting only a $25 per-meeting bonus.
Among 130 rejected postings containing monetary figures, 23.1% contained a missed pay disclosure (30 of 130, 95% CI 16.7% to 31.0%). The sample is restricted to rejections that hold a money figure and expose no salary field, because only those can move the rate. The 100 correct rejections were funding rounds, valuations, deal and ARR bands, sales quotas, referral bonuses, home-office and wellness stipends, learning budgets and tuition reimbursement. The misses were mostly ranges published with no label beside them, or with the label separated from the figure by a line of location text. This confirms that the detector deliberately trades recall for precision, which is why 44.9% should be read as a conservative observed estimate rather than a ceiling.
Misses cluster by applicant tracking system: Greenhouse 30.2% (16 of 53) against Ashby 9.8% (5 of 51), on the two systems with enough rejections in the sample to report.
No corrected rate is published. Scaling the miss rate up to the whole corpus would give a higher figure, but an honest correction would also have to subtract the false positives the same audit found, and carry the uncertainty of both rates rather than one. The result would be an estimate of an estimate. 44.9% is what was observed and can be recounted from the snapshot, so that is the number this report stands behind.
Language could not be stratified. The language field is unset on most listings, and the pool is overwhelmingly English, so these figures measure English postings. The drawn sample, every listing id and every verdict are kept with the snapshot so the audit can be recounted.
The four geographic categories, and the listings in them, are the same extraction the accessibility report publishes for the same date: 9,699 listings with a stated scope, split the same way. The two are not computed independently and then compared; this snapshot refuses to build if any of the four counts disagrees with the accessibility snapshot, so a reader opening both cannot find two different numbers for the same thing.
The headline rate is of all 19,563 active listings. The US comparison is of listings with a stated geographic scope: 6,189 naming the United States against 2,477 naming at least one country but not the United States, with unstated-scope listings excluded from both. Field shares are of the listings in that field. Employer figures are of 2,872 employers.
Currency is never converted: USD 4,324, EUR 51, GBP 35. Hourly, monthly and annual figures are never converted into one another. Only base pay enters a median; on-target earnings, commission and bonus are excluded. Where a pay period is not stated, an amount above $20,000 is treated as annual for the sensitivity table only, and that inference is shown rather than relied on.
A group is shown only with at least 100 listings, and a median only with at least 30 priced listings. Groups below those are omitted and named rather than shown with a caveat.
Salary ranges and pay labels are what employers published. The disclosure rate, the US gap and its ratio, the employer-level shares and every median in the sensitivity table are WFA calculations from those inputs, and are marked as such where they appear.
Every listing says which countries it can hire from before you apply. Browse by where you live or by field.