Fast, Open and Checked? A descriptive case study of an AI- assisted route to open publication

Generative AI has made it cheap to produce papers and equally cheap to produce plausible references. This paper reports a descriptive case study of one fast route to open publication: ten thematic papers, drafted with AI assistance from 81 newsletter articles by the author, and posted to Future Science Archives on 29 October 2026, the first at 20:40 and the ninth at 22:18 (UTC+7). The case is reported from the inside. The author recorded more than two dozen production errors found during drafting, including a retracted meta-analysis cited as evidence, reversed findings, and figures that differed from their sources. We then audited the reference lists of the ten papers. Parsing found 233 entries (about 198 unique works by a heuristic match). Of these, 65 (27.9%) were self-citations, 64 of them to the author’s own LinkedIn newsletter. Of the 168 external entries, 51 (30.4%) printed a DOI, 113 (67.3%) gave only a URL and 4 (2.4%) gave no locator. A printed DOI shows readiness for direct DOI lookup. It does not measure the coverage that any checking could achieve, and no DOI-registry or retraction-database lookup was run on the corpus. No assessment of whether sources support the claims made from them was undertaken. In a random sample of 25 external entries, AI-assisted web searches returned a putative match for 24 and displayed the cited locator for 16. The same AI system family helped draft the papers, and a search result shows that a work can be found, not that it supports a claim, so this is not an independent verification. The findings are descriptive. They do not establish that any verification approach is effective. We propose a minimum verification statement that documents responsibility, methods, coverage, unresolved cases and corrections, give a completed example, and specify a protocol for independent human assessment.