Search researchers by name or ORCID — with citation metrics, affiliations and research fields from the open OpenAlex dataset (CC0).
What this directory covers, and what it does not. It holds 32,325,772
author records, every one of them with three or more published works — measured 27 September 2026 on the
served table: not one record falls below that bar. 8,720,350 of them (27.0%) carry an ORCID, up
from 23.3% on 19 September, because in the eight days between, the ORCID population grew 21.4% and the directory 5.0%.
Until 19 September 2026 this directory held 7,369,024 records and said, correctly at the time, that it was
not everyone who meets the three-work bar, the open dataset then listing 30.8 million. The article and
author layers moved to our own server that day, the rest of those authors came with them, and that gap closed.
The 30.8 million was never a finishing line, and this notice should not be read as though it were: rebuilt on the
23 September 2026 OpenAlex release, the directory now holds 1,544,791 records more than it did on 19 September and
has passed that figure. The three-work bar is a rule, not a size, and the set it describes grows whenever the
source does.
Two coverage figures, and they count different things. Our journal–author graph holds
10,214,773 link rows, and those rows name 5,668,521 distinct authors. Of those,
435,538 (7.7%) have no profile in this directory behind them. Separately and from a different
source, 8,864,207 of the 32,325,772 records here (27.4%) are linked to no journal in our catalogue
at all. The two populations do not reconcile and are not built to.
Until 28 September 2026 this notice said the graph “names 10,214,773 authors”,
which is the link count and not an author count, and put the missing-profile share at 388,431 of 5,668,521 (6.9%)
with a caveat that author_coverage_mv rebuilds weekly and had not run against the new tables. Both
halves were wrong. rpc/author_coverage was already serving 435,538, and recomputing the view’s own
expression against the current tables returns 5,668,521 and 435,538 as well — so the view was never stale; the
page had stopped matching it. No automated check covers this figure, which is why nothing caught it.
Read an empty search result
as “not in this directory”, never as “not a researcher”. The gap was 2,549,835 (45.0%) until
30 August 2026, when the author table was rebuilt against the expanded catalogue, and 637,224 — 21% —
until 22 August, when it hid most of the best-cited researchers in the world. Fixed, and recorded as
errata E23.
Sorted by h-index by default, and that default is deliberate.
A very large citation count resting on a handful of works is the signature of one
name absorbing the work of many researchers — an author-disambiguation
failure in the upstream data, which we do not correct and cannot hide. Until 22
August those records held the top of the citation order outright: 14 of the top 20
had an h-index of 5 or lower. They no longer do, because the researchers who
should have been above them were missing from this directory and now are not
— the top 20 by citations today average an h-index of 188, and none is below
5. The same upstream failure has a second shape, though, and one of those is still
near the top: a name absorbing many researchers can also show as a hundred thousand
works under a two-letter name. The 30 August load brought in an “M.
Shoji” with 100,008 works and an h-index of 38, and it is still third by citations
two corpus rebuilds later. We show
it rather than remove it. The older artefacts sit further down; the largest of
them now ranks 95,956th by citations, one of ten records tied on the same total. Both orders are worth
reading, and neither is a ranking of people. Re-measured
28 September 2026 against the live API, over all 32,325,772 records. This said 198 and
63,984 until then — measurements taken 2 September 2026 on a 7,369,024-record table,
which then survived both the 19 September move to our own server and the 27 September
rebuild without anyone re-taking them; before that it said 225 and 36,120, from the
22 August table.
Source: OpenAlex authors dataset (CC0). Author records are algorithmically disambiguated by OpenAlex. The directory holds 32,325,772 author records, each with three or more works, and an ORCID on 8,720,350 of them (27.0%). It is not the most-cited researchers: it is everyone the source lists above the three-work bar, which since 19 September 2026 is the whole of that set rather than the 7,369,024 records drawn from our own journal–author graph before it. That set is not a fixed size: it grew by 1,544,791 records when the directory was rebuilt on the 23 September 2026 OpenAlex release, and 8,864,207 of the records in it are linked to no journal in our catalogue. The notice at the top of this page records what is and is not measured on the new table. Spotted a wrong record? Tell us.