Recently I read an article by Paul Kinlan of AI Focus about an experiment he did to capture a web page in flight and use AI to change it to match his whim. He calls it Interception.

Examples? Change the text of the article you are reading to a style you are more accustomed to. Change images in the article according to your preferences: removing or adding clothing, accessories or other attributes.

Kinlan approaches this idea from a tech perspective, which is fine and fascinating. But I immediately hooked onto what it would mean if it went mainstream.

Let me try to explain how such an approach would work. And why I bring up A/B testing.

Today, a website is designed, built and deployed to production. It’s a product, designed for a particular purpose and for a particular audience. It has a sender and a message, even if the message is marketing.

Many decisions are made before work is started, sometimes lengthy research is deemed necessary. Still, unknowns remain. That’s where A/B testing comes in, a trial and error way of optimising those last feet to the front door.

E-commerce is a great example, because you can easily measure what the difference in sales is between option A and option B.

With A/B testing users’ wallets decide which option wins. The chosen option is then set to default.

But what if users could use AI to hack into the information flow, and how would that work?

Picture the flow of information from user to the website and back to the user. The user presses a link in their browser, a message is sent to the website, the website sends information back to the users’ browser, the user views the information.

In Kinlan’s experiment, the browser would then pre-process that information. Browsing history could be used as source to train the AI, which could then change the incoming information according to the users’ liking. Rewrite text, refactor images, rewire the entire content. The user would view the information sent to them refashioned to their tastes.

Far more invasive than changing browser settings such as text size and fonts. Hyper-personalised content.

Think of our e-commerce website: endless disputes over pricing and conditions. Screenshots would become worthless.

Earlier this week I picked up on Kinlan’s article to discuss what such a feature would mean for the sociology of the internet, saying that internet users would be forever trapped in the Filter Bubble now reserved for social media.

Now I think it may be worse.

What use would A/B testing still be? You might add, what use would any design be? What use would it be to continue to carefully curate content?

We in the web business are careful to secure our websites from actors that would hack into our content and change it to their liking before it reaches the end-user. But if Interception went mainstream, they would hold office inside your device.