From Research to Publish in a Day — Co-Writing with ChatGPT and Skipping the Academic Gatekeepers

Reposted from my LinkedIn articles, first published August 12, 2025.

Today was supposed to be a writing day. My plan was simple: do some exploratory research on Chinese micro-dramas as groundwork for a module I’m teaching this autumn (2025).

The module’s focus is on getting students to think like creators and entrepreneurs. Micro-dramas felt like a perfect fit — something students could feasibly create themselves with real potential to develop into viable content businesses. I wanted to replace the current client brief (BBC Three — which feels stale and underwhelming these days) with something fresher, more entrepreneurial, and grounded in contemporary content ecosystems.

I already had a starting point: a 2024 industry report from China, full of deep-dive analysis on the economics, supply chains, and creative trends driving micro-drama. Step one was translation — ChatGPT helped break down sections and summarise key points. From there, I fed it additional literature, articles, and news items to build a broader context.

What began as a modest research exercise grew into something bigger. As I iterated with the AI, more interesting threads emerged — like the role of multi-channel networks (MCNs) in powering the micro-drama industry — and before I knew it, I was developing a coherent paper. By the end of the day, I had something that felt peer-review ready.

Lessons from Co-Writing with ChatGPT

ChatGPT doesn’t like long documents. Around the 3,000-word mark, it starts doing strange things — sometimes rewriting perfectly good sections. To avoid this, I learned to:

• Take regular snapshots of drafts I liked.
• Work in chunks of 1,000–1,500 words.
• Work to a structured outline so the through-line wasn’t lost.
• Regularly check in with ChatGPT to mitigate against task drift.

My role in the process went far beyond simply prompting ChatGPT and accepting whatever it produced. At every stage, I was making decisions about the shape, tone, and focus of the piece. I prompted for targeted analysis on specific themes, then compared the responses with my existing knowledge and the source material to judge accuracy and relevance. I fact-checked claims against original reports, academic articles, and trusted news sources, discarding anything I couldn’t verify. I decided which points to keep, which to cut, and where to expand, often querying ChatGPT for deeper dives on concepts I thought deserved more space — such as the role of multi-channel networks (MCNs) or the economics of short-form monetisation. I inserted references, restructured sentences for clarity, and adjusted emphasis to align with the argument I wanted to make. Where ChatGPT’s phrasing felt too generic, I rewrote it to reflect my own turn of phrase and rhetorical rhythm. In short, I acted as both editor and co-author, ensuring the end result was not simply an AI-generated text but a synthesis shaped by my critical judgment, thematic priorities, and stylistic choices. That’s why the article retains a discernible human voice, even though its polish and fluency are enhanced by the AI.

This wasn’t the first time I’d tried co-writing with AI, but it was the first time I felt I had something publishable at the end of it.

Why Self-Publishing Matters

I’ve written elsewhere about why I’m done with traditional academic publishing, but here’s the short version:

• It’s slow (often years from submission to publication).
• It’s exploitative — universities pay us to write, review, and edit, then pay publishers again to access that work.
• When I research, I find articles via Google Scholar, not journal homepages. Searchability matters more than prestige.

Today’s second experiment was to see whether I could take something I’d written — in this case, a piece co-produced with ChatGPT — and move it rapidly through a self-publishing workflow that would make it not only public but also searchable, citable, and unambiguously mine. The goals were straightforward but ambitious:

1. Produce a publishable preprint that, while not yet peer-reviewed, would meet the standard of clarity, coherence, and scholarly framing needed to be taken seriously.
2. Make it discoverable in the same way as a journal article, ensuring that it would appear in search results on platforms such as Google Scholar.
3. Assign it a persistent, credible citation through a DOI, so that anyone referencing it in future would have a stable link.
4. Establish clear authorship and provenance, ensuring it was attached to my ORCID profile and therefore integrated with my professional record.

By the end of the day, I had achieved all of this. The article was uploaded to Zenodo, instantly assigned a DOI, and automatically linked to my ORCID record. I mirrored it on OSF Preprints and ResearchGate, each pointing back to the Zenodo DOI. The paper is now live, citable, and — crucially — under my name, with a transparent record of its origin and version history. In effect, I compressed a process that normally takes months or years in academic publishing into a single afternoon, while retaining ownership and bypassing the gatekeeping structures of traditional journals.

The Self-Publishing Workflow

The workflow I tested today is deceptively simple, but the implications for academic independence are huge. It starts with the writing process itself, which in this case was a hybrid of AI assistance and human editorial judgement. ChatGPT provided rapid synthesis, structured drafts, and suggested language, while I took on the role of curator and editor — prompting targeted analysis, fact-checking against source material, refining the argument, and shaping the overall structure. This kept the voice mine, even while benefiting from the efficiency of AI.

Once I had a version I was happy with, I moved to the publishing stage. The first step was Zenodo, a free open-access repository backed by CERN. Uploading the PDF here gave me an instant DOI — something that normally requires going through a journal or institutional platform. The DOI not only provides a stable citation link but also gives the work more credibility in search results. Because my Zenodo account is connected to my ORCID profile, the publication was instantly added to my professional research record, visible to anyone searching for my work, including my own university. Zenodo also allows updates, so I can upload a revised version later — for example, after peer review — without losing the DOI or its search engine indexing.

The next step was to extend the reach by uploading the same document to OSF Preprints and ResearchGate. In both cases, I included the Zenodo DOI in the metadata and description, ensuring all three versions point to the same canonical record. This creates a web of discoverability — if someone finds the work on OSF or ResearchGate, they are still led to the Zenodo version, which remains the primary citable source.

Hanney, R. (2025). Micro-Drama: From Chinese Phenomenon to Global Trend – Preprint Version. https://doi.org/10.5281/zenodo.16798457

The result: within hours, my article was live in three different platforms, all linked by a DOI, already visible in Google Scholar, and permanently under my authorship. What normally takes months or even years in the traditional publishing model happened in a single day — and I retained full ownership of the work, the record, and the data trail.

The Peer Review Question

The one thing my experiment doesn’t yet include is peer review — that long-established marker of scholarly legitimacy. In the traditional model, peer review is bound up with journals, editorial boards, and publisher-controlled workflows. But there’s no inherent reason it needs to be that way. What I’m considering next is an open, invited peer review process for this paper, where colleagues and subject specialists can provide structured feedback, and those reviews are then published in full alongside an updated version of the article.

This approach has two important effects. First, it makes the process transparent — readers can see exactly what was said, what was changed in response, and how the paper evolved. Second, it makes scholarship iterative — rather than being frozen in a final ‘version of record’, the work remains a living document that can be revised and improved over time. With Zenodo’s versioning, each update keeps the DOI lineage intact, so citations still resolve to the most recent form, with all earlier versions archived and accessible.

If more of us embraced this model — peer-reviewing each other’s self-published DOIs — we could begin to detach peer review from the commercial publishing apparatus entirely. The review process would remain rigorous but be freed from the artificial scarcity and access barriers of the paywalled journal system. In Foucauldian terms, we could start to prise open the ‘regimes of truth’ that define what counts as legitimate knowledge in academia, redistributing that power among scholars themselves.

Done collectively, this could be genuinely liberating: enabling faster knowledge circulation, ensuring work is accessible from day one, and letting the scholarly community own both the discourse and its record.

… and in the end

In one day, I took an idea from rough notes to a citable, publicly accessible preprint — complete with a DOI, linked across multiple platforms, and ready for open peer review. It’s a small but tangible step toward reclaiming scholarly publishing for scholars themselves.

I’d be interested to hear what you think: Could you see yourself using this model? Would you be willing to peer review work in an open, public way? Let’s start that conversation.