PRECISE: checking what AI says against what its sources actually say

From: PRECISE <precise.research.protocol_at_nyob>
Date: Wed, 9 Sep 2026 01:12:49 -0700
To: CODE4LIB_at_LISTS.CLIR.ORG
Hello everyone,

I’m Sima Cohen, the creator of PRECISE, an AI research verification
workflow.

Over the past year and a half, I’ve been mapping errors in AI research
outputs and running many manual tests. One recurring problem is that a
citation can lead to a real source while the accompanying sentence changes
its meaning: a finding becomes broader, a date disappears, or a statement
is attributed to the wrong organization.

That seems particularly relevant when AI is used for reference work,
literature reviews or research involving collections.

I built PRECISE around checking claims against their sources and preserving
details such as scope, dates and attribution through to a deliverable
tailored to the user’s needs.

A free, limited PRECISE LITE demonstration is available through the website
<https://precise-research.teachable.com/p/precise-protocol?utm_source=code4lib&utm_medium=community&utm_campaign=precise_lite_2026&utm_content=code4lib_post_01>
.

I’d be interested to hear how others here check whether an AI summary
preserves what a source actually establishes.

Sima Cohen
Received on Wed Sep 09 2026 - 04:13:09 EDT