Content will be judged by quality not volume - but no-one has told AI

The inevitability of AIGen content spreading isn't inevitable at all, as any glance at real world uses tells us.

by Rob Corbidge

Published: 15:09, 08 May 2025

When there is a clear divergence between the general above-the-line narrative and the observable facts, then it makes any predictions about the near future extremely difficult to make.

This is the case with our own industry and the thieving hydra that is GenAI.

A video of a recent talk given by the CEO of Cloudflare, Matthew Prince, has been doing the rounds this week. In it, Prince uses an interesting metric of pages scraped by a search engine versus the number of visitors that the site scraped would receive in return. It's a metric that makes sense coming from him, especially when you consider where Cloudflare sits in the general tech stack. The news Prince gives isn't good. Using his metric, we've a 2:1 ratio of pages scraped to visitors ten years ago, to a ratio of 6:1 now, mainly due to Google's increasingly intrusive and irritating attempts to keep us on pages it owns.

That's not even the bad news. Prince says that the scrape-to-visitor ratio for OpenAI is 250:1, and other GenAI systems are even worse.

Speaking as a grizzled veteran of a number of tech hype wars over the past few decades, it's a truth that the one around GenAI is the biggest ever seen. The astronomical sums of money being thrown around and into things that make a loss, the utterly ridiculous claims around capabilities that are just around the corner, always just around the corner, and worse, the widespread view that the end justifies the means when it comes to where data for GenAI use is sourced from.

There's obviously a fear many people have in that by criticising the assumptions of the noisy AI bandwagon, they risk looking like a 15th century monk telling his team of tonsured scribes that illuminated manuscripts are the only reliable medium and that this newfangled printing press thing is just a flash in the pan.

The theoretical monk would presumably make an argument of quality versus quantity, and you can see how the AIGen content versus original content fits into such a template. Yet, in the current paradigm, the quantity is being sourced from the quality, as original content is used by GenAI to provide information with limited reference to the sources it's drawn from. That might be fine for the most anodyne of search queries, but as one delves deeper, any remotely active mind will question the information being provided with "says who?".

We are also in the position of such GenAI systems being trained more and more on synthetic data, so we are getting a synthesis of a synthesis. We all know processed food isn't good for us, and now likewise, the same can be said of processed content.

The physicist and controversial populariser of science, Dr. Sabine Hossenfelder, recently explained what she sees as the shortcomings of GenAI in her work of reviewing scientific papers. The issue is the GenAI has no qualitative parameter for actual content, even if its output makes grammatical sense. Every scientific paper is "interesting" and "good" according to ChatGPT, even if it's not, even if it's deeply flawed. GenAI has no critical reasoning, all it produces is conformity, and that will come as little surprise to those that understand how LLMs actually work.

So, are we setting our face against a technological inevitability? Are we advocating still making our tools from stone when iron has become available? Not at all. To even compare the source of original content, the human brain, to the aggregation engines that produce AI content is absurd. It's actually insulting. Yet the prophets of AI have no qualms about such comparisons. There is nothing inevitable about the particular progress of a technology, even if it seems like it. There's a divergence in what we are told the future is going to be, and what is observable on the ground.

On the one hand, publishers are anxious not to be outflanked by GenAI, and many are suffering a severe case of FOMO, leading to questionable decisions. On the other, we seem to be on the verge of the Great Breaking-Up of Meta and Google, and no one in publishing is sure where the pieces will land afterwards. It's important we don't just swap one Tech Overlord for another.