Pay transparency, measured: what 9,188 live job ads actually say

Job hunting advice tends to assume that a salary is something you negotiate. Before you can negotiate it, someone has to tell you what it is, and mostly nobody does. We can put a number on that, because we hold the postings.

This is a count of our own corpus: 9,188 live postings from 518 employers across 51 countries, gathered between 2026-07-10 and 2026-09-25. Of those, 2,206 carry a salary figure we could read, which is 24%. A further 642 mention pay somewhere in the text without our parser catching a figure. So the real rate is somewhere in 24% to 31%. Call it a third at the most generous reading, and three quarters of the market silent.

Where you are decides, not who you are applying to

Split the same question by country and it stops looking like a matter of taste.

• United States: 2,635 postings, 69% to 78% state pay.
• Canada: 230 postings, 47% to 63%.
• United Kingdom: 648 postings, 19% to 26%.
• France: 456 postings, 0% to 5%.
• Germany: 1,624 postings, 0% to 1%.
• India: 1,426 postings, 0% to 1%.
• United Arab Emirates: 195 postings, 0% to 2%.

The United States is not culturally more generous with information than Germany. It has California, New York and Washington, each of which requires a pay range in the advert itself. Canada has British Columbia. Those laws exist, and the disclosure rate in those two countries is an order of magnitude above everywhere else in this corpus. Nothing else in the data separates them.

The EU deadline has already passed

This is the part worth sitting with. The European Union's pay transparency directive carried a transposition deadline of 7 June 2026, which was 4 months before this count was taken. The three EU countries where we hold enough postings to measure are Germany at 0% to 1%, France at 0% to 5% and the Netherlands at 0% to 1%.

A deadline passing is not the same as a job advert changing, and the two should not be confused. Transposition happens at a different speed and with different teeth in every member state, and we have not read the resulting national laws, so we are not going to tell you what any of them now requires. What this count can say is narrower and still worth saying: several months past the deadline, the advert itself has not started carrying the number in these three markets. If you are applying in Europe, check what your own country now entitles you to, and ask for the band either way.

The people with the least information are the ones with the least power

The sharpest number in the whole count is not a country. Of 946 live internships, 1% to 2% say what they pay. Of 353 apprenticeships, 0%. Compare that with 28% to 36% for ordinary jobs.

The candidates with no salary history, no benchmark and the least ability to walk away are the ones told least about what the work is worth. That is not an oversight in a minority of postings. It is how nearly the entire category is written.

Remote work goes the other way, and for a mechanical reason: 1,128 remote postings run at 39% to 47% against 22% to 29% for 8,018 onsite ones. A remote role can be seen by an applicant in a state that requires the range, so the range goes in. One jurisdiction's law leaks across every border the posting crosses.

Law sets the floor. Employers still choose.

The tempting conclusion is that this is entirely legislation and no employer ever decides anything. Our own data says otherwise. Narrow to United States postings from employers with at least 40 live roles, so the jurisdiction and roughly the sector are held constant, and the spread is enormous.

• Datadog: 212 postings, 92%.
• Databricks: 456 postings, 89% to 97%.
• Anthropic: 495 postings, 87% to 98%.
• Figma: 101 postings, 86% to 100%.
• GitLab: 84 postings, 56% to 74%.
• Stripe: 329 postings, 6% to 8%.
• Jane Street: 101 postings, 0%.

Two of those are quantitative trading firms competing for the same graduates. Optiver states pay on 68% of its 57 United States postings. Jane Street, across 101, states it on 0%. Same city, same candidates, same law, opposite answers. Whatever the reason, it is a decision somebody made, not a constraint somebody inherited.

How we counted, and where this is weak

Three things you should hold against these numbers.

• This is our ingested corpus, not the labour market. It leans towards employers who publish structured postings on their own career pages, plus Germany's federal job service. A market of small firms hiring through WhatsApp is invisible here, and that market is large in India.
• Our parser misses real disclosures. One posting in this corpus states an annual range of 265.000 to 310.000 euro and still parses to no figure, because the thousands separator is a full stop. That is why every rate above is a range rather than a single number: the low end is what we read, the high end assumes every posting that mentions pay in prose really did disclose it.
• Percentages on small samples are noise, so no country under 100 live postings and no employer under 40 appears above.

Everything here is produced by one script in our repository against the live table, so the same count can be run again and will move as the market does. That is the only reason we are comfortable publishing it. A number nobody can recompute is a claim, and this product exists because claims stopped being worth much.

What to do with this if you are the one applying

• Assume silence is the default, not a signal about the specific employer. At 24% to 31% disclosure, filtering out every ad without a range mostly filters out the market.
• Ask in your first reply, in one line, before any interview is scheduled. In the countries above that sit near zero, nobody will find the question unusual, because everyone is having to ask it.
• Anchor on the role rather than on your last salary. Where the employer will not publish a band, your own evidence of what you can do is the only lever you control. That is worth more than a number you cannot see.