After several months of development, we are excited to present a new plugin for OJS 3.3/3.4+ called IssueSpotlight AI. This tool integrates the Google Gemini API (Free Tier!!) directly into the platform to transform static metadata into analytical and visual layers.
The goal is to take an issue’s metadata (titles, abstracts, and affiliations) and generate four types of automated AI analysis that provide high editorial and analytical value:
Intelligent Editorial Synthesis: Acts as an Editor-in-Chief, drafting a thematic narrative of the issue.
Innovation Radar: Identifies emerging topics and trends through a visual “Packed Bubble Chart”.
SDG Impact (Agenda 2030): Evaluates how the research aligns with the UN Sustainable Development Goals.
Global Institutional Map: Advanced geolocation of authors with AI-powered normalization of university names.
Efficiency and Zero Cost (Free Tier)
We have designed the plugin to be sustainable and cost-effective:
Built for the Free Tier: It runs on the Gemini 2.5 Flash Lite free tier (approx. 20 requests/day).
One-Time Analysis: The analysis is triggered once by the editor. All LLM responses are stored in a custom OJS database table, so subsequent access by readers does not consume additional requests or API quota.
Privacy: Only public metadata (titles/abstracts) is sent to the API. No personal or unpublished data is ever shared.
Live Examples
See the plugin in action in our OJS, Revistes UPC:
Thank you very much for the suggestion. I’ll review the instructions you provided and implement the necessary changes to release the plugin in the Plugin Gallery.
We tested the plugin on two issues of CHIMIA, the Swiss Chemical Society’s long-standing chemistry journal, which publishes topical issues with review articles. Each issue also has a community part that serves the Swiss chemical community and divisions.
Vol. 78 No. 6 (2024): Sustainable Development Goals in Chemistry in Switzerland, https://www.chimia.ch/chimia/issue/view/2024_06 . The scientific articles were labelled by the authors with the SDGs, therefore we can assess the quality of the AI SDG analysis.
Conclusion
Due to weaknesses both in design/technical aspects and in the AI results, we think that the plugin is still in a very experimental state and needs overhaul. For our journal, currently it does not provide added value to the reader, but rather degrades the quality of the carefully edited issue. Therefore, we will not pursue it for some time.
I keep the analysis of the two issues for some time so that others can make up their mind.
Detailled feedback from me and two other editors of the journals
Design/technical aspects
there are a lot of hard-coded Spanish terms (e.g. “ODS”, “Tendencia”) in the templates that need to be replaced with locale keys.
Is there any order in the author list below the map?
AI analysis
Editorial Summary:
We did not look at the Editorial Summary closely (and would not need it), because the guest editors already provide their own Editorial. Content-wise, the AI summary grasps the main aspects of the issues, but the language is often clumsy. You may compare the editorials of the two issues with what AI provided. The guest editors texts are more fluent, address the reader directly, provide broader context which AI does not, and have a personal touch AI simply can’t match.
“I also find a lot of the AI information that it has picked up to be quite inaccurate and then how would we control this?”
Global map
“The global map is typically going to centred on CH and I do not think is going to be very useful to the journal.”
“The map may or may not have a better distribution depending on the issue. What I miss in the map is an additional layer of collaboration relations between institutions, which would provide added value. Also, just a list of authors (the good old author index) without links to the articles is not helpful.”
Innovation radar
“{…} (analogue of a word cloud) - I do not find anything here that I cannot glean from browsing through the contents list and reading the abstracts. Since the terms in the graphic are not linked to specific articles in the issue (at least I cannot see links), I do not find this very helpful.”
“Furthermore, what are the relations between the bubbles? I can drag one and they get rearranged, but I can’t see any forces (usually driven by relation strength) that would position them in a meaningful way. Also, what are confidence measures of match of the selected terms? This is a difficult topic as I have learned from a recent project {…}”
Interestingly, the keywords AI finds often differ from those that the authors provided in their articles (IMO, this is not to meant negatively, just an observation by me)
SDG Impact
“This is perhaps the most revealing analysis, and at least it comes with more information than the ‘innovation radar’. However, it’s noteworthy that the SDGs cited by authors in issue 6, 2025 are not mirrored by the analysis. The AI analysis does not appear to have picked up the SDG logo information from the articles. I am unclear how, for example, Gender Equality is measured: AI seems to be focusing on one article (10th anniversary of Women in Natural Sciences). I would have expected AI to look at e.g. an analysis of genders of authors but this is not foolproof - not all names are clearly identified with gender.”
“Out of the 16 assigned SDGs, the AI detects only 5 and even doesn’t fit the distribution correctly.”
Details of the analysis (article count = count of articles the authors have assigned this SDG label).
GitHub Repository: IssueSpotlight AI