Overview
Publication Tracker, Appetence’s open-source (meta)data repository for publishers, has a connector to Amazon Bedrock that runs a prompt on a model the publisher chooses, with guardrails, inside the publisher’s Virtual Private Cloud. Appetence reported in October 2026 that a manuscript scope check that uses this connector is already in production with publishers. When a manuscript is submitted, its abstract is compared with the journal’s scope, and an alert appears in Publication Tracker if it does not fit.
Appetence also offers AI-assisted review through the same connector, from technical screening, such as checking whether a submission matches the journal’s guidelines, up to what it describes as “a full, substantive peer review”. Appetence says the publisher decides how far the review goes, while Appetence selects the model and sets the guardrails. Reviews can run automatically or on demand.
Appetence says Publication Tracker data can also be used with Amazon Quick, Amazon’s AI assistant for business intelligence, so that a publisher can ask questions of its data in plain language and generate charts and dashboards from it.
Appetence, the technology partner for the STM Integrity Hub, is developing an AI agent in Publication Tracker to carry out the manual steps a publisher’s research integrity team takes after submissions have been screened. Ivo Verbeek, director of Appetence, described those steps in August 2026 as interpreting the signals that screening tools produce, analysing the submission further where needed, and handing the case to the journal editor, who makes the editorial decision. According to Verbeek, the agent can carry out each step automatically and unattended: it decides when further analysis is warranted, pulls up other relevant information, interprets the content and turns the result into a recommendation for the editor. He says Appetence’s work is concentrated on grounding, the mechanism intended to keep the agent accurate and reliable so that editors can trust its recommendations.
In November 2026 Verbeek described screening submitted manuscripts with Claude connected to Publication Tracker through its MCP server, which retrieves a submission’s metadata, extracted metadata and full text with a single prompt. He suggests running this as a scheduled task each morning, so that editors start the day with the screening results. He also set out instructions for an AI agent that follows new submissions to a journal, runs a scope check, technical screening and journal guideline check, and recommends immediate rejection, acceptance or return to the author for revision. Once the desk editor confirms a return, the agent manages the correspondence with the author from the desk editor’s mailbox until the manuscript is ready for desk acceptance, which the desk editor also confirms.
Notes and quotes
- “The only difference is the prompt, the model and the guardrails.” Appetence, on how technical screening differs from a full AI review, October 2026.
- “Put simply, it learns from previous cases and always operates under full control, with a human in the loop.” Ivo Verbeek, Appetence, on the research integrity agent, August 2026.
- “Ultimately, however, it is the human desk editor who makes the decision.” Ivo Verbeek, Appetence, November 2026.
- Verbeek says his November example of agent instructions is simplified and leaves out some necessary safeguards.
Sources
- Publication Tracker: Where Your Publishing (Meta)Data Meets AI, Appetence Newsletter, LinkedIn, 1 October 2026
- Lean Editorial: Doing More with Less in Research Integrity, Ivo Verbeek, ALPSP blog, 25 August 2026
- Lean editorial: doing more with AI (connectors and tasks), Ivo Verbeek, LinkedIn, 4 November 2026
- Lean editorial: doing more with AI (agentic desk editor workflow), Ivo Verbeek, LinkedIn, November 2026
This entry was written with the help of AI, from the sources listed. Please check the sources before relying on it, and let us know if anything needs correcting.
