Skip to the page
AtlasCord, home

How it works

AtlasCord draws who researchers work with from their published papers, and who trained whom from what people and universities confirm.

What it is

AtlasCord is a living map of academic lineage and collaboration. Each researcher sits at the centre of their own map: supervisors above, students below, academic siblings beside, and co-authors around them. The map can be replayed year by year, and opened out to institutions, countries and the whole world.

It began as a pilot at the University for Development Studies in Ghana, and it is built for researchers, graduate students and universities anywhere.

Tools built on your record

  • CV studio: build promotion and funder CVs (UG, UDS, West Africa, NIH, NSF, UKRI, Europass) straight from confirmed papers, history and indexing journal counts, with PDF, Word and frozen share links.
  • Figure studio: turn tables, CSV or Excel into publication-ready statistical plots, math graphs, PRISMA 2020 flows and 90° engineering block diagrams with exact statistics, draggable layout, colorblind previews and vector SVG/PDF export.
  • Draft check: verify every DOI and reference against Crossref and OpenAlex with a Verification map, review paragraph-anchored writing notes and similarity checks, state AI use transparently, and share specific versions with a supervisor.
  • Interactive discovery maps: explore matching experts on a Field map, inspect cited sources on an Evidence map in Ask the map, and see warm-introducer paths in Build a team.

Your first few minutes

  1. Sign in with Google or a link sent to your email. There is no password.
  2. Is this you? You see the author records in the public record that may be yours, and confirm the ones that are. Your papers, co-authors and the institutions printed on them arrive.
  3. Your map appears, already drawn with your co-authors.
  4. Add your history: your degrees, posts, supervisors and students. Type it, or upload a CV for the map to read; you check every item before it is saved.

Where the data comes from

WhatSource
Papers, co-authors and citationsOpenAlex, released under CC0. You confirm which records are yours; new papers wait for your answer and are never added silently.
Institutions and where they areROR, released under CC0.
Degrees, posts, supervisors and studentsYou, and your university's records where it attests them.
Links to your profiles elsewhereYou. An ORCID iD, Google Scholar or ResearchGate address is kept as a link and never fetched. Google Scholar and ResearchGate are never read.

Figures such as citation counts are worked out from the papers you confirm, so they read lower than Google Scholar, which counts more kinds of document. Your profile says so.

Every line on the map is drawn to show how far it can be trusted:

  • Public record: co-authorship from published papers, with the source shown.
  • Confirmed: both people agreed.
  • Attested: the university's records back it.
  • Unconfirmed: visible only to the person who added it.

A supervision link is never guessed from papers or shown to anyone else until the other person confirms it, the university attests it, or a confirmed co-supervisor or two confirmed fellow students vouch for it.

People who have not joined

A researcher in the public record who has not joined appears only to signed-in members, with their name, their published papers and the institutions printed on them, marked as not a member. They are never shown to search engines or to visitors who are not signed in.

Anyone can see what is held about them, without an account, and from there stop being shown or be removed. A removed person is never imported again.

What the AI does

The map works out the facts. A language model only words them: the answers in Ask the map, the reasons in a proposed team, a summary of what changed in a year, and a draft of the history it reads from an uploaded CV. Its words are checked against the facts before anyone sees them, and nothing it drafts is saved until a person confirms it. Model calls run on our servers, never in your browser. With no model set up, every feature still works with plain sentences.

No decision about a person is made automatically. Matching a request to a record only decides how soon a link is sent; anything else is decided by a person.

What it does not do

  • No feed, no messaging and no advertising.
  • No email to someone who has not joined, unless a member chooses to send one. Even then it is one email, in our name, with a permanent opt-out.
  • No physics simulation: the same data always draws the same picture, so your map looks the same tomorrow.