AI-Attributed Shortfalls In Search Traffic Are Everywhere, But They’re Disproportionately Affecting Fashion

This week, Wikipedia has reported an 8% fall in visitors year-over-year, and the root cause they identify shouldn’t surprise anyone who’s been paying attention to the way AI chatbot apps (and the models that power them, trained on essentially the entire internet) have been steadily swallowing the experience of searching the web and interacting with sources.

“Search” might now be synonymous with “Google”, which has leaned more heavily into keeping people inside its platforms over time, but the original definition was a straightforward one. Search engines were intended to match a user with the content from across the vastness of the web that best matched their query, and the same user would then visit that page, completing a value exchange that largely left users satisfied, that funnelled some revenue to publishers, and that turned Google – as basically the sole intermediary – into one of the best businesses, and richest monopolies, on the planet.

AI search is changing that pattern, by crawling sources either during training or at runtime, using web scrapers, and then serving up answers that, historically, users would have needed to click to find. Ideally those answers include citations that at least give the user the option to investigate the source – but the data shows that this is a poor replacement for a traditional link when it’s judged by established metrics.

As one of the world’s biggest websites is demonstrating: the web we’ve known is in marked decline. And while a <10% shortfall in a single year might not read like much, next year’s statistics are likely to show an even more pronounced version of the same trend – especially with this week’s announcement of ChatGPT Atlas, the new web browser from OpenAI. (All but certain to come complete with a different definition of “browsing” to the one we operate under today.)

This is an area where The Interline, as a free-to-read web entity, has been an open book. As part of this analysis, we updated our understanding of our own traffic: our audience continues to grow, year-over-year (a 23% uplift from 2023-24 to 2024-25) through a range of different means, but we, too, can observe a clear pattern: organic search referrals are down 16%, while referrals from AI services, primarily ChatGPT, are up from essentially zero in 2023 to close to 6% of all traffic today. And this does not account for the impossible-to-track scenario where our reports, web content, and other pieces are cited in AI responses but users do not then click through.

But as significant as this shift is starting to feel for the wider content and publishing industry, new research from TANK, a UK digital agency that specialises in search and performance data, suggests that its impact could be even more pronounced in fashion.

Tracking more than eight hundred companies across sixteen sectors, and comparing traffic to their web properties before and after Google introduced its AI-powered Overviews feature, the research revealed that, across the board,  growth in organic traffic fell from twenty-six percent to four. But for fashion businesses this slide was much more pronounced: from a higher bar of plus thirty-three percentage to minus three. 

And in absolute terms, the number of fashion pages ranking in Google’s top results dropped by more than twenty percent during the survey period. The only industry experiencing a more pronounced decline in search-directed traffic was hospitality.

TANK’s data doesn’t separate fashion media outlets and publishers  from brands and retailers, but the pattern makes intuitive sense when you consider how much fashion content online tends to follow familiar formulas: “how to wear,” “what to buy,” list-based guides for coming seasons, marketplace listings, and PDP pages for eCommerce that follow a very predictable formula. That kind of repetition likely makes fashion easier for both readers and algorithms to digest, and it may also make it easier for summarisation systems to reproduce the same ideas with minimal new input. Studies of large language models show that they rely on patterns and sentence templates when condensing or generating text. And it’s also quickly become the case that companies selling online have put brevity of content first and foremost in their product detail – recognising that consumers rarely read it, but also inadvertently distilling their own content in a way that makes it easier for AI models to understand and repeat… and then also potentially using those same AI models to write the next churn of content.

The drop in clicks also does not mean that the  audience has disappeared. There are, on aggregate, more searches and more interest flowing through the system than ever – that attention is simply being captured before it reaches the brand or retailer.

Martin Harris, Head of Digital at TANK, put it this way when The Interline spoke to him for this analysis: 

“AI search is redefining discovery for fashion businesses, shifting visibility from keyword rankings to credibility. Success now depends on orchestrating intent, structuring ideas so AI can understand, contextualise, and confidently surface them.

The fundamentals still hold. If you rank well in traditional search, you’ll likely fare well in AI search, because authority, clarity, and expertise carry across.

What’s changed is the journey: discovery is no longer linear but continuous. As AI fuels an always-on loop of intent, trusted content keeps earning visibility beyond the first click. In a visual, reputation-led industry like fashion, originality and authority matter more than ever, regurgitated opinion won’t cut through when algorithms prioritise distinct perspectives and authentic expertise.”

In a practical sense, Harris is referring to a long-predicted shift in the purpose of creating web content: from writing for a human reader, to writing and designing for an AI model to read.

To underline that idea, other research published last week revealed just how quickly AI content has grown as a proportion of all web pages: from less than 20% at the time ChatGPT was launched, to more than 50% today. But, ironically, that glut of AI content should – at least in theory – actually be serving to push authoritative, human-created content to the top of the pile.Google’s own documentation nods in that direction, saying its systems look for content that shows experience and authority – with the fundamental difference being that those systems will now be capturing that content in AI overviews, rather than sending readers to the underlying pages.

Of course, discovery is only one piece of the story. The same technologies unsettling it are also being built into the day-to-day operations of fashion. In that context, the technology doesn’t just absorb attention, it redirects it, optimising the more practical work that makes the industry more efficient. 

Wikipedia’s data shows what happens when information still travels but loses its link back to the source. Fashion’s deeper impact demonstrates how quickly a universal trend can become especially acute for a particular sector. And what we see in publishing might be less discouraging, but still has important parallels for our readers, who have always been in the business of being found – but who now need to quickly reckon with a substantial change in what that means.