Start here
Where to begin depends on why you came
Truvace files every measured gain and harm from AI under a topic, with the study or report behind it. Pick the reader closest to you.
I came as
Route 01 of 04
Researchers
You want the studies on one question, and a way to cite or correct them.
Next route: Policy staff and buyersFind your topic
Open its count: how many records found a gain, how many found a problem and which are strongest.
Cite a record
Each record has a permanent ID (like TRV-2026-2474) and a ready-made citation on its record page, next to a public history of every change.
Ask for a review
If a record covers your study and we got something wrong, or we missed a study, tell us. We email you back with what we did.
Download the data
The Index exports every matching row as CSV or JSON.
2,375 published recordsCSVJSON
Route 02 of 04
Policy staff and buyers
You need to know what is established before a decision, and how strong it is.
Next route: TeachersOpen a topic
See the count on each side and a table of how strong that evidence is.
Diagnostics & Imaging455 records279 show a gain92 show both84 show a problem17 of 42 topics have enough records on both sides to weigh.
See pooled results
Figures from systematic reviews, each one quoted from the review it came from.
Compare 10 AI companies
The documented gains and problems tied to each one.
Read how scoring works
Including which inputs an AI model sets and which are fixed formulas.
- 40% Evidence-backed gains
- 40% Evidence-backed problems
- 20% Evidence quality and breadth
Route 03 of 04
Teachers
You want real, sourced cases a class can argue about.
Next route: JournalistsStart with Traces
A Trace is one study or story that found a gain and a problem at once. They make the best discussion cases.
Use Talk it over
Open any record for four discussion questions, with that record's gain, problem, evidence and caveat filled in under them.
- This record found a gain and a problem in the same place. Which matters more to you, and who might answer differently?
- How do we know? What can this kind of evidence show, and what can’t it?
- What does it leave open? What else would you want to know before acting on it?
- Have you seen AI used this way where you study, work or live? Did it go the same way?
Browse Education
Records on tutoring, cheating, classrooms and how students study.
Read the Reading of the Month
One longer essay a month, built from that month's records.
Route 04 of 04
Journalists
You want what is new, and the primary source behind it.
Next route: ResearchersSee the latest records
Newest first. Or follow the RSS feed at truvace.com/rss.xml.
Follow a record to its source
Every record links to the paper, filing or report it came from. A company's claims about its own product are never counted as evidence.
- 10 peer-reviewed
- 5 journalism
Among them: Acta Neurologica Belgica, Surgical Laparoscopy, Endoscopy & Percutaneous Techniques, The Guardian, Journal of Stomatology Oral and Maxillofacial Surgery, JNCI Cancer Spectrum.
Get the Monday brief
One email with the week's new records.
105 records published
- 16Thu
- 17Fri
- 18Sat
- 12Sun
- 13Mon
- 15Tue
- 14Wed
Days are counted in UTC. The last bar is today so far.
Send us a story
Tell us what we missed, with its source.
ace