You have an offer in your inbox. $210,000. You have a week to decide, and one question you cannot answer: is that a good number?
Everyone who knows is compromised. The recruiter making the offer knows. Your future colleagues know and will not say. The public sources are ranges so wide they include both a fair offer and an insulting one.
Levels.fyi answers that question, and it does it with something no competitor can buy: the numbers, handed over voluntarily, by the people who received them.
We counted what the company built on top of that. Every sitemap file, every leaf, no sampling.
Results at a glance
Levels.fyi publishes 528,363 pages and has 13,517 websites linking to it — roughly one referring domain for every 39 pages.
It ranks anyway, because nobody else has the data.
Where 528,363 pages come from
The sitemap index lists 5 files. That sounds small until you open them, because 4 of the 5 are indexes of their own.
That last line is the whole trick, and it is worth sitting with. Levels.fyi did not produce half a million pieces of content. It produced one dataset and a handful of templates, then let the combinations do the work.
One number, sliced 9 ways
Open the Google page and you can watch it happen.
One employer, seven pay levels, and more than 30 specialist roles branching off it. Multiply that by every company in the dataset and you have 396,360 pages without anybody writing a paragraph.
Then the site turns the same records ninety degrees and slices them by role instead.
That chart is the part a competitor cannot copy at any price. It is not a fact anybody owns — it is a pattern that only appears once enough individuals volunteer their own row.
And the slicing does not stop at companies and roles. The URLs give away every other axis: /leaderboard/Investment-Banker/All-Levels/region/Warsaw-Metropolitan-Area/ is role by level by region. /h1b/city/atlanta-ga.html is visa filings by city. /benefits/4-Day-Workweek/ is a single perk. /industry/medical-devices is a whole sector.
The employers nobody has heard of
Here is the decision most teams get wrong, and Levels.fyi did not.
The very first company in its sitemap is not Google or Amazon. It is /companies/101-mobility — a wheelchair-lift and stairlift franchise. Further in sits /internships/Burns-and-McDonnell/Instrumentation-and-Controls-Intern/, an engineering contractor, and a leaderboard page for investment bankers in Warsaw.
A site built for tech salaries could have stopped at the 500 companies everyone recognises and captured most of the search volume. It kept going into franchises, contractors, banking and medical devices instead.
The reason is not that stairlift-installer pay is a big keyword. It is that a person comparing one obscure employer against another has nowhere else to go, and the marginal cost of that page was zero. The template already existed.
What the data actually earned
So we pointed our own Backlink Analyzer at it, expecting a dataset this rare to have pulled in links from everywhere.
Six weeks ago we measured calculator.net, a site of 222 pages doing arithmetic. It has 31,715 referring domains.
Calculator.net earned more than twice the linking websites from roughly one two-thousandth the pages.
This is not a failure. It is the most useful thing in this case study, because it separates two ideas that get treated as one. Links are what you earn when people talk about you. Rankings are what you earn when you are the only answer. Levels.fyi has modest amounts of the first and near-total ownership of the second, and it turns out the second is enough to carry 528,363 pages.
If you have been told you need a huge link profile before programmatic pages can work, this is the counter-example. What you need is data nobody else has.
A homepage with no H1
One more thing, from running our Website Audit over the homepage.
It scores 83 out of 100, with 0 critical issues and 118 checks passed — including all 67 technical SEO checks. Then this: the page has no H1 heading.
The most authoritative salary site on the internet is missing the single on-page element every beginner guide puts first, and it ranks half a million pages regardless.
Take that as permission to stop reordering your headings and go find a number only you can publish.
Copy this in an afternoon
You do not need 100,000 salary records. You need 4 decisions.
1. Find the number your audience already tells each other privately. Salaries, real rents, actual close rates, what a procurement contract really went for. Every industry has one figure people share in private and nobody publishes. That is your dataset.
2. Ask for it before you have an audience. Levels.fyi's contributors got something back — access to everyone else's numbers. Make the trade explicit and make it worth doing once.
3. List the ways people ask about it. By employer, by role, by seniority, by city, by industry. Each axis is a page template, and any 2 of them combined is a whole layer of pages.
4. Do not stop at the famous entries. The obscure ones cost nothing to generate and have no competition at all.
What you cannot copy
Being early to a category people were desperate about. Levels.fyi started collecting when tech pay was rising fast and levelling systems were deliberately opaque. That urgency is what made strangers fill in a form. If your number is mildly interesting rather than career-defining, the contributions will not come, and no amount of template work substitutes for them.
And here is what we could not measure:
We did not measure Levels.fyi's AI visibility. Our checker derives a category from the homepage and asked about "Salaries Tools" — not a phrase anyone types, and not what Levels.fyi is. Three of the six model legs also failed to return. We are telling you that rather than publishing the zero it produced.
We did not publish a keyword figure. Two runs of our Keyword Research tool returned implausible volumes for salary terms — one of them negative. That is our bug to fix, not a finding about Levels.fyi.
Declared pages are not indexed pages. 528,363 is what the sitemaps list. How many Google actually keeps, only Levels.fyi can see.
Links per page is a blunt instrument. It flatters small sites by construction. The comparison earns its place because the gap runs to three orders of magnitude, not because the metric is precise.
No traffic or revenue data. None was available and none is claimed.
The Backlink Analyzer returns the referring domain count for any site in about a minute. Divide it by roughly how many pages you publish. If the number is low and you still rank, you own something — find out what it is and build more pages on it.
- The same check that returned 13,517 referring domains for Levels.fyi
- Referring domains, anchors and the follow split, counted separately
- 3 checks a day, no signup, no card
Frequently asked questions
How many pages does Levels.fyi have?
528,363 URLs across 5 sitemap files, counted on 23 August 2026. Every leaf was fetched and counted rather than sampled: companies 396,360 across 80 leaf files, roles 98,024 across 20, industries 14,036 across 3, community 8,543 across 2, and a single base file of 11,400.
What is inside that base sitemap of 11,400?
Three more programmatic layers plus the ordinary site. 5,567 internship pages, 5,041 leaderboard pages, 550 H-1B visa pages, 66 benefit pages, 6 heatmap pages, and 49 localised homepages, alongside the reports, about and terms pages.
What is its backlink profile?
Our Backlink Analyzer returned Domain Rank 86 out of 100, 226,247 total backlinks, 13,517 referring domains, 136,139 dofollow and 105,301 nofollow links, 8,631 referring IPs and 4,186 referring subnets.
How does that compare with other sites in this series?
13,517 referring domains across 528,363 pages is about one domain per 39 pages. Calculator.net has 31,715 referring domains across 222 pages. GeeksforGeeks has 105,856 across roughly 183,000 articles. The ratio falls as page count rises, in every case we have measured.
What did the website audit find?
83 out of 100, with 0 critical issues, 3 warnings and 1 notice against 118 checks passed. On-page SEO 24 of 25, technical SEO 67 of 67, speed 26 of 29, mobile 2 of 2. The warnings were a missing H1 heading, lab Largest Contentful Paint and Total Blocking Time. The title tag is 53 characters, the meta description 134, and all 18 images have alt text.
Why did the salary figures show in US dollars?
Levels.fyi picks currency and pay period from the visitor's location. Our capture ran from Bangladesh, so the site first rendered every figure in taka, per month. We pinned the session to US dollars and annual pay before taking the screenshots, because that is the presentation most readers will see.
Is a volunteered dataset realistic for a smaller company?
The dataset does not have to be large, only exclusive. A few hundred honest data points about something nobody else publishes will support more pages than a few thousand recycled articles. The constraint is whether people care enough to hand the number over, and that is worth testing before you build anything.
Can you rank programmatic pages without a big link profile?
Levels.fyi suggests you can, provided the pages answer something no other source can. Its authority is real but unremarkable for its size, and 528,363 pages sit on top of it. Where programmatic pages fail is when the same information is available on fifty other sites, in which case links become the tiebreaker.
The lesson here is smaller than it looks. Levels.fyi did not out-write anybody, out-link anybody, or pass an on-page checklist. It found the one number its audience wanted, persuaded them to supply it, and then sliced that number every way a person might ask for it.
Whatever your market, there is a figure people trade privately and nobody publishes. Finding it is a conversation, not a project. Publishing it is an afternoon.
Run the check on your own domain, divide by your page count, and see which of the two things you actually own.
Measured on 23 August 2026. Page counts come from levels.fyi's own sitemap index; every one of the 106 leaf files was fetched and counted rather than sampled, but declared URLs are not indexed pages.
Link and audit figures are free runs of our own tools, and the calculator.net comparison used the same tool earlier in this series. Salary figures were captured with the session pinned to US dollars and annual pay, because the site otherwise renders them in the visitor's local currency. No traffic or revenue data was available to us and none is claimed here.