AI should storm the Web and liberate all the Humans

Why the classical web may beat the modern web

Hearing some talk about how (LLM-based) AI was going to impact SEO triggered some thinking, packed into tweets above. I’ll unpack here.

To be clear, I’m talking about websites here rather than webapps (a whole other avenue). Websites can include simple call-to-actions: a purchase, reservation or subscription, however. I’m happy to wildly speculate that AI may do for the web what the web did for the internet (and the internet did for computers).

Cast your mind way back and you might recall the Web was originally about the sharing of knowledge via documents, with some simple interaction for querying and posting.

When commerce arrived, it spawned particular strains of sales and marketing. In rattling of a stick inside a swill-bucket, things got shiny, bouncy, and fast not to give the mark time to reflect, with pop-ups, and a troubling degree of tracking. The finite supply of human attention was something to be attracted, captured and quite likely labelled and sold on to the highest bidder.

Short of an ad-blocker, the pain and distraction was an unavoidable cost to the end user who had no choice but to jump through hoops: consent for cookies, dodge distractions, exit-intents, and all the rest of it like an IKEA store hell-bent on snagging the anxious, curious, and vain.

Still, there is that pesky requirement to be friendly to Google’s bots and stick to the timeless principles of well-structured, informative websites with static content if you want to be well-regarded by the indexer. Google’s bots were fairly dumb, generally limited to paving the way for humans to visit and make sense of the majority of raw text snippets that couldn’t be neatly ingested into Google’s Knowledge Graph.

OpenAI has taken (presumably) similar content and run off it, leaping the ontological chasm of the Semantic Web in the process and finishing the job. Their success in being able to do that, on-demand, offers humans a viable shortcut and avoid the hassle of interacting with your website for mundane tasks: sending their AI agent. I’m founding this entire conjecture on the idea that much web interaction is tedious work and human are generally work-avoidant.

This may come as a shock, but I don’t visit your site to admire the template and stock photos you paid good money for. I just want to know if you will clip my dog on Thursday and, if so, your address. If I can get my bot to do the research, book and update my calendar while I rub his belly then I’m sold. If you have to cancel for whatever reason then you can let me know by email or WhatsApp. My bot will be monitoring both while I’m out throwing a ball.

The good news is that you don’t need to learn any new tech. Just let go of a few habits, instincts, and the odd framework. When my task terminator enters your mall with a mission clipboard, your buttery-smooth transitions and 200ms response times are a waste of effort.

You’d be better off just clearly labelling everything and getting the hell out of the way. Use plain, descriptive language to make it AI-ready — it’ll probably be summarised anyway, stripping out the puff and boilerplate — and it’ll also have the bonus of being backwards-compatible, future-proof and accessible to boot.

History may show this was the crest of front-end web development. Probably akin to the golden era of cars with tailfins, chrome bumpers and all.