Why a website you never touch again becomes invisible to AI

Recent research into how AI assistants choose their sources shows something worth noticing: the vast majority of cited content has been updated within the past year. Content that sits untouched for longer gets used less and less often, even when the information itself is still correct.

What that means in practice

A familiar example: a webshop put up a page three years ago with opening hours, the product range and delivery terms. Nothing has changed since, even though the range has actually grown and delivery times have shifted. For a visitor who happens to land on that page, that is already frustrating. For an AI assistant deciding which source to cite, it is a reason to skip the page: more recent, more active sources weigh more heavily.

Why this goes beyond a technical issue

This is not about a page that is broken or throwing an error. The page works fine, loads fast, and looks exactly as it did three years ago. The issue is in what an AI model concludes from that: a page that never changes gets treated as a source that is less current, and therefore less reliable, than a page that is visibly maintained.

What this actually means to do

For a small business the fix is rarely a full rebuild. It is about a handful of pages that really matter, such as the services page, pricing information or the most common questions, and going through them regularly to update what is current. Adding a seasonal example, refreshing a figure, folding a new customer question into the FAQ. That is not a big job, but the signal it sends counts toward every question an AI assistant answers afterwards.

Read more about how AI assistants pick their sources in the explanation of AEO.