When The Economist publishes two articles, one of them a leader, about the same issue you know it needs to be taken seriously. In its edition of 19 July 2025, one of its leaders is titled To survive the AI age, the web needs a new business model, and a longer article in its Business section states AI is killing the web. Can anything save it? Both are well worth the effort of creating an account to read them. The leader states crisply that “the danger is that, as answer-engines take readers away, they are removing the incentive for content to be created,” concluding that “if nothing changes, the risk is of a modern-day tragedy of the commons. The shared resource of the open web will be over-exploited, leading to its eventual exhaustion.”
The problem lies in what it so appropriately refers to as Google’s change from being a search to an answer engine, a subject further explored by the Pew Research Center’s timely report of their study summarised here by humans. That demonstrates that Google “users are less likely to click on links when an AI summary appears in the results”.
These themes are central to my previous account of PageRank and plagiarism. As with others who publish original work on the web, I’m used to sites that copy entire articles, such as MacMegasite. Within an hour of its publication here, for example, that had stolen the whole of Friday’s Mac article, word for word.
Note the sponsor ads on the right confirming this theft was performed for profit, and in flagrant breach of copyright. Yet Google’s answers are perfectly comfortable using stolen content and even providing links to it.
What neither The Economist nor Pew’s research have considered is how inaccurate AI-generated content is corrupting the web and diluting the output of humans, with the flaws in its wholesale plagiarism. I inadvertently exposed this with my article Boot disk structure in macOS, iOS and iPadOS, and AI cryptexes, where those structures were summarised in diagrams that LLMs currently don’t access. As a result, several different AI summaries completely missed the main points, and concentrated instead on the textual notes I supplemented those diagrams with.
These aren’t the ‘hallucinations’ now seemingly tolerated as unusual aberrations of AI, but signs of a more general failure that can’t be disguised behind a euphemism. The truth is that, while AI does have its uses, it can only be as useful as the human knowledge on which it depends, and its provenance.
Nothing that I publish here has come from AI or answer engines. Every word that is written comes from this human. Where there are errors or mistakes, they are mine and not the product of an LLM. I do not use AI such as Writing Tools for proofreading, to summarise articles, or to write comments. The same applies to content that I write for publication elsewhere, including my regular sections in MacFormat and MacLife magazines, whose publisher explicitly forbids the use of AI to produce content.
The time has come to stand up and make it clear that none of the content of this blog is prepared with the aid of Writing Tools, Apple Intelligence, Large Language Models, or Artificial Intelligence. What you read here is researched, composed and corrected by a human.
To make this distinction clear, this site will now display an owl emoji. Owls have long and deeply ingrained associations with the night, and with wisdom and learning, going back to the classical civilisations of the Mediterranean.
Hendrik Goltzius shows a classical and fairly complete set of this Roman goddess’s attributes in his Minerva (as the Personification of Wisdom) from 1611: the owl denoting wisdom and learning, her distinctive decorated helmet, a spear, books, and beauty. The artist also anticipated the arrival of AI over four centuries later, in the figure in the left background, King Midas with the ears of an ass.
Each article published here now appear under the mark 🦉 No AI content to make this clear, and I encourage others who write blogs and online content free from AI to do likewise. It’s time to stand together.