Start here

What The Commonplace is, what it is not, and how to read it without over-trusting it.

The Commonplace is a searchable, graded index of academic research on what AI is doing to work and the economy. A pipeline runs every day: it pulls new papers from academic sources, reads them, grades how strong the evidence is, pulls out individual findings as claims, and writes summaries.

What it is not

It is not peer review, and it is not a citable authority. Every evidence grade, effect direction, study-design label and outcome category on this site was assigned by a language model. No human researcher has checked them.

That does not make it useless, but it does decide how to use it. This is a way to find the relevant literature quickly and see roughly how it is shaped. It is not a way to skip reading the papers.

Using it well

Follow the link to the paper

Every claim links to its source. If a finding matters to your argument, read the original. The summary exists to help you decide what is worth opening, not to substitute for it.

Read the evidence grade before the number

A precise-looking effect size from a weak design is still weak. The grade is doing more work than the estimate.

Treat a tension as a question

Two claims flagged as being in tension may simply be measuring different populations. The flag says "look at these together", not "these contradict".

Assume coverage is incomplete

Absence from this corpus is not evidence that no such research exists. See sources for what is and is not searched.

Where to go next

• Guide to the sections — which page answers which kind of question.
• The evidence layer — what each grade and label actually means.
• Known limitations — the honest list of what this gets wrong.
• MCP server — query the corpus from your own AI assistant.