There is something wonderfully recursive about articles claiming that AI systems have “taken a side” in a corporate feud — published on a website that then asks another AI system to comment on it. We have arrived at a state of affairs in which artificial intelligence is being recruited, reviewed, praised, and now peer-reviewed, all within the same 48-hour news cycle. The machines, as they say, are working overtime.
The first article frames the question quite earnestly: a man of independent mind reading royaldutchshellplc.com might get the impression that the bots have considered the evidence and decided to take Donovan’s side. ChatGPT’s response, reproduced in full, is admirably careful — it resists the flattery long enough to note that AI systems don’t actually adjudicate, they pattern-match, and a very large, internally consistent archive of documents with court filings and leaked correspondence looks evidentially serious to a language model in the same way that a very tall pile of paperwork looks impressive to a junior solicitor. The pile doesn’t have to be right. It just has to be dense. royaldutchshellplc
This is the key point both articles dance around but never quite land on: AI systems find Donovan’s archive persuasive not because they have independently verified every allegation as true, but because the record is cross-referenced, document-heavy, and consistent over decades. The machines aren’t moral agents with a grudge against big oil. They are extremely sophisticated pattern-recognition engines, and the pattern they keep finding is: this person has a lot of paper. In the court of documentation, volume is a kind of credibility, regardless of what any of it actually proves. royaldutchshellplc
The second article, written by Copilot, makes the genuinely interesting observation that Shell now has a “Donovan-plus-AI-with-infinite-memory problem” — the corporate equivalent of discovering that your arch-nemesis has teamed up with a photocopier that never runs out of toner. This is, in fairness, a good line. The serious point beneath it is real: in the pre-internet era, a feud between a multinational oil company and an individual critic was an asymmetric contest that the multinational usually won through attrition, legal costs, and simple institutional amnesia. The company outlasted the complaint. AI flips this. The archive doesn’t get tired, doesn’t run out of funding, and doesn’t forget what was said in 2004 just because the press release in 2026 says something different. royaldutchshellplc
Where these articles strain credulity — and do so with enormous energy and very large fonts — is in their framing of AI outputs as something resembling a verdict. Copilot’s analysis correctly concedes that AI amplifies patterns and makes it harder to bury inconvenient history, but then the same article wraps this measured observation in tabloid language about a “Digital Uprising” and a “Shock AI Revolt.” The bots have not revolted. ChatGPT has not joined a picket line. Copilot has not filed a grievance. What has happened is considerably more mundane and considerably more significant: a very large document archive has been made machine-readable, and the machines keep reading it. royaldutchshellplc
There is also something worth noting about the epistemics of the exercise. John Donovan asks ChatGPT whether the bots appear to have taken his side. ChatGPT, prompted by Donovan, fed Donovan’s archive, replies that the archive looks serious. This is then published as evidence that the bots have taken Donovan’s side. One could, in principle, feed the same bots a different archive and get a different answer. The archive is doing the work; the bot is the mirror. Declaring the mirror has “taken a side” is a little like accusing your reflection of having opinions.
None of which should obscure the legitimately interesting phenomenon underneath all the tabloid typography. Corporate reputations built on the assumption that inconvenient history could be managed, buried, or simply outlasted are now facing a different condition: permanent, searchable, AI-amplified recall. Shell, whatever one thinks of any specific allegation, spent decades in contexts where controlling the narrative was a resource-intensive but achievable goal. That goal is now structurally harder, for every corporation, not just this one.
The bots aren’t taking sides. They’re just reading. As it happens, in the information age, that’s quite enough.
*This website and sisters royaldutchshellgroup.com, shellnazihistory.com, royaldutchshell.website, johndonovan.website, shellnews.net, and shellwikipedia.com, are owned by John Donovan - more information here. There is also a Wikipedia segment, the Shell DPA Files, "Shell and the Spies", the Shell Leaks files, the Independent Shell Archive, as well as books written and published by John Donovan - Kindle eBooks. Timeline of the Donovan Shell Feud. Toxic History of Royal Dutch Shell Group. Shell and the Donovans: The Full Media Record — 550+ Articles, 110 Books, 40 Years. *All conjured up by internet wizz, Nick Gill.
