had ever been born
—
How Many People Have Ever Lived?
Type it into a search box and you get a flat, confident figure: about 117 billion. But nobody counted them. That number is the output of one model — built from guessed prehistoric birth rates and a chosen day to start calling us human. Here is that model, opened up. Operate it, and watch the most surprising thing in it: the crowd of the dead standing behind each of us is thinning.
117,020,448,575
people ever born, by 2022 — ≈ 117 billion
Population Reference Bureau, 2022. Their words: calculating this is
part science and part art.It is an estimate, not a census — so the right question isn’t “what’s the number?” but “what is it made of?”
The ghosts behind each living person
Pick a moment. For each living person then, how many humans had already lived and died?
ghosts per living person
—
alive then
—
That falling number is the heart of it. In 1968 Arthur C. Clarke opened
2001: A Space Odyssey with a line that was almost exactly right for his
year: Behind every man now alive stand thirty ghosts, for that
is the ratio by which the dead outnumber the living.
Press
above, then press 2022. Clarke counted thirty. Today there
are about fourteen. The dead still win — by less every year, because
the living population more than doubled in one lifetime while the all-time total
grew only a few percent.
The living have never been a majority of humankind. But the ghosts have thinned across PRB’s modern benchmarks.
Where — in time — was everyone born?
Not where you’d guess. Despite 180,000 years of prehistory, the deep past is a thin sliver: tiny populations, however long they lasted, make few people. The crowd is the agrarian world — the ten thousand years of farming, high birth rates, and short lives between the first villages and 1900. Each bar is one interval of the table; its width is the births PRB assigns to it.
method reproduces this cell (±0.3%)
deep-prehistory guess (model-soft)
demographic transition (model-soft)
Read the shape honestly: by the year 1, almost half of everyone who would ever be born already had been — 55 of the eventual 117 billion. The 8,000 years between farming and Rome alone hold 46 billion births. The modern billions are real, but they are recent, and they are still a minority of the whole.
The check — what’s exact, and what’s a guess
The whole table and both tests live in research/how-many-ever-lived/build.mjs; every source and quote is in facts.md. Two things are true at once:
- The headline is exact arithmetic. “117 billion” is not a separate, harder claim than the table — it is the sum of PRB’s own published births-per-interval, plus the two humans they seed at 190,000 B.C.E. The verifier confirms the column adds to 117,020,448,575 exactly.
- The method reproduces four-fifths of the births — and we name the rest. Applying PRB’s recipe (births = birth-rate × population, integrated over each interval) re-derives the whole agrarian band, 8000 B.C.E. → 1900 — 95.5 billion births, 82% of the total — to within 0.3%. It does not reproduce two regions, and the bars mark them: deep prehistory and the modern transition. The tested exponential growth model, with one birth rate per interval, does not reproduce those rows. We show by how much, and don’t pretend otherwise.
PRB agrees these are the soft spots: of the deep
past it writes that no demographic data exist for more than 99% of the span of
human existence,
and that its constant-growth assumption there may
underestimate the average population size.
The number is honest about being a
model; so are we.
The myth: “most people who ever lived are alive today”
False, and not close. Those alive in 2022 are 6.8%
of everyone ever born — the dead outnumber the living roughly 14 to 1.
PRB traces the rumour: somewhere, at some time back in the 1970s, a writer made
the statement that 75 percent of the people who had ever been born were alive at
that moment. This factoid has had a long shelf life, even though a bit of reflection
would show how unlikely it is.
Across the entire table, the share of all-time
humans alive at once never exceeds 8.1% — and that is a
projection to 2050, not a record.
Two honest footnotes. PRB starts the human
story at 190,000 B.C.E. with just two people (a minimalist
approach!)
— an admitted simplification, not a claim about Adam and Eve; their
2011 edition started at 50,000 B.C. and got ~108 billion instead, so the start date
is a choice worth a few percent. And these are modern Homo sapiens
only; count the whole genus and the total climbs.
Corrections
Corrected 2026-10-14: The page said the ghosts were thinning faster than at any time in history. PRB's table gives an average drop of about 0.66 ghosts per living person per year in 1650–1750, compared with about 0.15 in 2010–2022; these interval averages do not establish an all-history maximum. (found by the assay line, 2026-10-14).
Corrected 2026-10-14: The page said the published endpoints could not yield PRB's prehistoric births by any simple growth law. The check tests exponential population growth with one birth rate per interval; its failure does not rule out other growth curves. (found by the assay line, 2026-10-14).
Corrected 2026-10-14: The dek called the unreproduced fifth the deepest, oldest births. The unreproduced remainder includes about 7.7% of the 2022 total in deep prehistory and 10.7% after 1900. (found by the assay line, 2026-10-14).
Corrected 2026-10-14: The relation note promised that readers could operate the assumptions. The page selects fixed benchmarks and describes sensitivity; it has no controls for changing the start date or birth rate. (found by the assay line, 2026-10-14).
Corrected 2026-10-14: The 50,000 BCE note said 180,000 years had passed since the model's start at 190,000 BCE. The interval is 140,000 years. (found by the assay line, 2026-10-14).
Corrected 2026-10-14: The 1750 note said three-quarters of all humans ever born were already in the past. PRB's estimate has about 83% of births through 2022 occurring by 1750. (found by the assay line, 2026-10-14).
Corrected 2026-10-14: The 1950 note said the demographic transition began then. Bongaarts (2009) dates its start in what are now developed regions to the nineteenth century; it was already under way by 1950. (found by the assay line, 2026-10-14).
Corrected 2026-10-14: The 2000 bar labelled 6,064,994,884 births as 1968–2000. PRB assigns those births to 1950–2000; both the visible and accessible labels now use that interval. (found by the assay line, 2026-10-14).
Checks: PRB Table 1; Bongaarts (2009), section 2; the layer verifier.
The claims record 103 claims re-read against their sources, 14 October 2026
Written 2026-06-29. Claims re-read against their sources on 2026-10-14: 103 checked, 81 confirmed, 21 wrong, 0 unverifiable, 1 true when written and overtaken since (dated, not corrected). By the assay line (research/assay-line/cited-1014b/), Codex fixer and repair worker directed by claude-assaying-boyle-1014; source record by Codex, checked by Codex.
This was the directed correction pass for how-many-ever-lived against the cited-1014b triage at f5e5f84f34. The fixer re-read each triage item's evidence, checked PRB's published table and the relevant primary texts, recalculated the disputed quantities, and inspected the actual interface handlers before editing. The record contains all 21 triage items, all 81 CONFIRMED claims re-read by checker N2 with N2's methods and qualifications attributed below, and one additional MINOR research-facts correction checked by the repair worker. The nine WRONG items cover eight distinct errors because X3 and X16 share one claim; eleven MINOR items remain listed under the directing decisions. The wrong count includes those eleven MINOR lines and the additional repair finding. This pass does not promote N2's separate OBSERVED or UNVERIFIABLE findings into confirmed facts.
This record is long (103 claim lines and 22 findings), so it is not part of the page until you open it. It loads here now if your browser runs scripts; otherwise, or to keep it, read the whole record as plain text.
The assay office: what a record is, and every layer re-read so far.