Alchemist Review, from Hum and Grounded AI, assists editors and reviewers with deep manuscript analysis, citation validation and an interactive chatbot. IOP Publishing has expanded its agreement to scale the tool across most of its proprietary journal portfolio, with the partnership running into 2026.
Benefits & outcomes
Early manuscript assessment with less editorial workload.
Citation validation built in.
Reduced reviewer fatigue and faster turnaround.
Scale: Hum says Alchemist Review processed more than a million manuscripts in the past year. Some large publishers turn it on for even a smaller portfolio and hit hundreds a day.
Runs novelty and impact assessments (effectively literature reviews) that feed live, in-production routing decisions at the front door.
Notes & quotes
IOP's expansion announcement: https://blog.hum.works/posts/iop-publishing-expands — ongoing expert input from IOP editors shaping new features.
Also: AIP Publishing partnership for quality control on AIP Conference Proceedings, and an IEEE pilot — https://www.sspnet.org/community/news/ieee-launches-pilot-with-hums-alchemist-review/
Editor testimonials from a July 2026 IOP slide: "I love the Alchemist Review because it's saving me a lot of time. I've been able to handle more manuscripts than I usually do." (Editor) — "So far it is working spectacularly as a papermill triage tool (like 50% better hit rate than some of the other things we are playing with)." (Publishing Ops) — "I use Alchemist Review for every manuscript. It aligns well with my own assessment, and sometimes catches rigor or clinical issues that would have taken me much longer to find." (EIC) [Source: slide in Helen's screenshot archive, 15/07/2026]
Purpose-Led Publishing pilot (AIP Publishing, APS, IOP Publishing; from April 2025, one journal team per publisher) published findings: real value in citation checking, metadata validation and ethics compliance, but the manuscript digest "didn't always capture a paper's nuance" — helpful for structured checks, limited where interpretive judgment is needed. Findings: https://www.purposeledpublishing.org/post/peer-review-assistant-trial — analysis in C&E's Brief: https://www.ce-strategy.com/the-brief/humain-peer-review/
From the PurePub.ai "AI Reviewers" panel (14 May 2026), Dustin Smith (Hum): multi-agent under the hood with Gemini as the "workhorse"; Hum frames the front-door role as editorial triage/routing rather than replacing peer reviewers — "we're not gonna hot-swap peer reviewers for robots." Claims no hallucination-related negative feedback (editor thumbs-down) in the past year, treating in-the-wild expert feedback as the gold-standard eval. Does not train its fine-tuned models on customer data ("written into all our contracts"); improves the harness, not the models. Internally frames the system as a "colleague," not a "tool."
Publisher / company: Hum (in collaboration with GroundedAI)
Date of mention: 11/03/2025
Source: https://blog.hum.works/posts/hum-launches-alchemist-review