A world crawled by machines
Cloudflare CEO Matthew Prince confirmed on 3 June 2026 that bots now generate more traffic than humans across the entire web. We had already sensed this by watching social media, but it now extends well beyond social networks. Anthropic’s bot crawls 10,300 pages for every single referral it sends to publishers. GPTBot stands at 903 to 1. Every website, every page, is constantly crawled by machines whose purpose is to build an all-encompassing knowledge of the world. In doing so, these systems obviously do not read your site the way we would: rather, we could say they consume it as raw material for building answers. In practical terms, this means that for a machine, certain kinds of graphic elements, rhetorical constructions and layout matter only so much. What matters is the hierarchy of information. The Economist has decided to respond to this reality head-on. According to Digiday, the magazine is experimenting with versions of its articles designed to be read by AI agents alone: content structured specifically to be processed, cited and summarised by AI systems, running in parallel with the editorial version intended for human readers. Two tracks, two architectures, two objectives. A single site. A brilliant answer to a problem that has not yet surfaced, and one that raises a key question for brands in the AI era: how do you govern your presence in an ecosystem where the audience is no longer human?Who will have the courage to do the same?
The truth we must now come to terms with is this: in 2026 a brand’s website does not serve only the customers who visit it. It serves the AI systems that crawl it, index it and use it as a source to answer users’ questions. Whenever someone asks an AI system what it thinks of a company, which is the best option in a category or whether a brand is trustworthy, the system builds its answer from what it has found by crawling the web. However, since, as we have seen, the optimal structure for a human reader is not the optimal structure for a bot, engagement and storytelling serve no real purpose. A document written for an AI must produce citations and nothing else. And that is enough. GEO (Generative Engine Optimization) is the answer the industry has begun to build. The Economist has chosen a more radical answer: not optimising the existing site for both audiences, but building separate architectures for each.What now?
The choice made by The Economist leads to a clear conclusion: the website is no longer a communication tool aimed at humans alone. It is infrastructure that must serve two kinds of audience with different needs. And the further AI advances, the more this will hold true. This does not mean every organisation should build two sites, obviously, but asking what an artificial intelligence would take from a given page has now become necessary. Because perhaps, very soon, speaking only to humans will be the fastest way to stop attracting other humans.Domande frequenti
Won’t AI systems also use the human version of the site?
They will use it, but less efficiently. An AI system that has to extract precise information from a narrative article written to engage a human reader must do far more interpretive work than one that finds the same information in a structured format. In a world where AI systems account for 57.5% of web traffic, optimising for them is no longer an optional technical choice: it is a condition of visibility.
Who should be responsible for a brand’s presence in artificial intelligence systems?
It cannot be the sole responsibility of the SEO team. Visibility in generated answers depends on the site’s architecture, but also on the clarity of positioning, the consistency of published information, the external sources that describe the company and the ability to make products, expertise and claims verifiable. That is why Bliss approaches GEO as an issue that cuts across online positioning, content and Brand Governance: the goal is not simply to make a page more readable to a crawler, but to govern the full body of information from which an AI system will reconstruct the company’s identity. On a web that machines also frequent, the question therefore becomes the same one that applies to any other touchpoint: who is deciding what the brand communicates when it is not the brand speaking directly?
What does it mean in practice to have a version of your website optimised for AI systems?
It means having pages with a precise data structure (Schema.org markup), key information organised so that it can be easily extracted without reading the entire article, claims backed by explicitly cited, verifiable sources, and a consistent presence across all the platforms AI systems use as primary sources: Wikipedia, LinkedIn, leading news outlets. It does not necessarily require a separate site: it requires the information architecture of the existing site to be designed with two types of reader in mind instead of one.