Generative AI search results are reducing online traffic to sector websites

As new digital metrics are devised in response to AI-powered searches, we have an opportunity to reconsider how both performance and purpose are tallied. We can frame responses to AI-powered search in terms of purpose, rather than performance.

Google’s Artificial Intelligence (AI) Search Overview feature – the AI generated answer you now get at the top of any search you put into Google – has now rolled out in more than 200 countries.

Appearing at the top of search engine results pages, overviews are AI-generated responses to user queries. The answers provided are probabilistic – meaning Google’s AI produces replies that are statistically most likely to be helpful, based on its training data, live web pages, and as of 2026, social media content.

The launch of Google’s AI Overview (and competitor equivalents) has not been uncontroversial. Early versions provided some users with incorrect information – including drastically wrong health advice and facts that are wrong (sometimes referred to as AI “hallucinations”). There are also quite a few legal battles against Google pertaining to anti-trust of its overview feature, defamation, and copyright infringement claimed by publishers. Nonetheless, AI overviews are increasingly popular.

Google reports that overviews appear on approximately 50% of all their searches, engaging two billion users per month. The tech giant also claims that overviews do not reduce click-throughs to content, despite contradicting evidence.  

Overviews have significantly reshaped internet search experiences. It quickly became apparent that they would disrupt the substrate of digital economies – clicks. By cultivating “zero-click” searches – when user queries are sufficiently answered by the overview alone – these products can disincentivise visits to websites. By some estimates, web search traffic has fallen by 58% since Google’s Overviews launched. For sites dependent on click-throughs for revenue generation, this is existential. Accordingly, a new market logic has emerged, in which an appearance as a cited source within the overview text becomes the priority.

Cultural organisations may have noticed the shift too. While not typically dependent on click-through rates for revenue, many Galleries, Libraries, Archives and Museums (GLAM organisations) assess website performance by converting traffic into Key Performance Indicators (KPIs). However, as documented in a 2025 report by One Further, search traffic to sector websites has decreased since overviews launched in 2024.

“Informational content” has been particularly impacted – with, for instance, queries about visiting hours, biographies and collections all potentially answerable via an overview; providing of course, the source content being drawn on is correct and deemed authoritative by the search engine.

Zero-click engagements therefore challenge typical growth-based metrics used to evaluate website success. Yet, if a GLAM website correctly informs an overview – ensuring its content appears prominently and outcompetes poor quality information sources – user needs are still being met. So while traffic decreases might feel confronting, the emergence of AI overviews gives us pause to consider what it is we’re counting.

For the purposes of reporting, KPIs will need to be revised. Organisations may be forecasting some reductions in web traffic, while aiming to increase overview citations – with Google Search Console now providing reports detailing webpage visibility in overview responses.

But beyond reconfiguring metrics, the sector is also afforded an opportunity to rearticulate our websites (and by extension, social media) not only in the language of performance, but in terms of purpose.

Growth-oriented metrics always risked serving as crude yardsticks. While necessary to gauge how effectively digital platforms serve audiences and deliver strategic goals, at its worst, arbitration by growth alone omits the responsibilities our platforms have always had as constructive contributors to an often-unstable information ecosystem.

This is not a new observation. In addition to contributing to the breadth of operational activities, from revenue generation to reputation management, GLAM websites and social media have long played a vital role in shaping how, where and by whom cultural content is encountered. This is because digital practice not only shapes how cultural organisations disseminate knowledge and experiences, but also creates knowledge and experience in its own right.

Understanding the epistemic effects of our online activities is part of the public service delivered by digital teams across the sector; expressed through good accessibility practices, representative content strategies, critical assessments of biases or omissions found within cultural datasets (including those used in the training of Large Language Models) – and much more beyond.

Digital practitioners are also required to take critical stances when adopting or responding to evolving technologies, such as AI and its market logics, given our decisions may tacitly boost, normalise, discourage, or delegitimise widespread usage.

Yes, the emergence of AI overviews necessitates a redraw of our performance metrics, and – akin to the shift to search-engine-optimisation-informed digital strategies that occurred circa 2014 – amendments to our ways of working. However, the transition to a post-overview internet also returns us to more fundamental considerations of purpose. Because, as ever, how GLAM organisations represent their work digitally remains consequential to who gets to experience it – whether or not the encounter is tallied in a click.

Hannah Barton is the head of digital media at National Museums Scotland