And the final speaker at the SEASON 2025 conference, very fittingly, is our host Dirk Lewandowski, presenting findings from a study commissioned by the German State Media Authorities that focussed on the transformation of online search by the introduction of artificial intelligence technologies. This is of course still in progress, so no definitive results should be expected yet. The study was conducted in May 2025, and results will be published in mid-October.
But we can already reflect on whether the introduction of AI represents a revolutionary transformation, or merely an incremental change; and on how AI-enhanced search affects the economic models of content producers. The study examined this for conventional search engines, dedicated AI search engines, and AI chatbots.
Generative AI enables the generation of human-sounding text, and the summarisation of information objects; but it suffers from hallucinations, and this underlying problem may not be solvable given its design. AI in search takes the dialogue between the information seeker and the information system to a new level; information seeking and information use are converging as a result. With AI, a search engine no longer finds and links to existing information objects, but in fact creates new information objects in response to a query.
This does not simply replace conventional search engines, however; rather, AI is being integrated into such search engines, or search is integrated into AI chatbots – both models may converge into one within the coming years. Through this process, at any rate, the ‘task frontier’ shifts: from finding through learning and investigating towards creating, and possibly further yet. This does not decrease the volume of search queries, however – rather, since the introduction of AI into search (or search into AI) the range of uses for these technologies, and therefore the number of queries, has increased steadily.
Previously, search engines intermediated between users and external information objects; AI responses generate new information objects that are based on but not simply extracted verbatim from the underlying documents. Retrieval-augmented generation is used for this process of generating AI responses, but this also means that such newly generated information objects can present diverging perspectives on a given topic. This can result in inaccuracies, too.
Users have often attributed even conventional search results to the search engine, rather than the sources it drew on; this trend is very likely to accelerate further with the presence of AI-generated information objects. But importantly, in this new scenario the search engine will also become liable for the content of these new objects.
This also affects user attention and traffic. With the placement of AI Overviews at the top of Google results, they dominate user attention, and click-throughs are known to follow visual attention. With search engine traffic still playing a major role in traffic to most Websites, this is likely to result in a considerable decline in user traffic to Websites (such as news publishers), significantly impacting on their economic bottom line.
And AI answers are not mere summaries of the top search results: they are fan-out queries expanding on the original query, whose results are then summarised. Clicking on the sources provided for AI summaries will not necessarily lead users to the information presented in these summaries, therefore.
A further implication of these changes is that the search engine’s data store (or index) is likely to change: where search indices previously contained all content available on the Web, now the inclusion of other sources, and the opting out of some content producers from search engine use, will affect then make-up of this data store. And opting out is complicated: it is possible to block Web crawlers, and block crawlers from using Web content for AI training, but not to block the use of content in AI content generation.
Content producers will be affected by this in different ways: content producers that monetise Web visitors directly will be negatively affected at an economic level; content producers that have other funding sources may seek to have more of their content included in order to gain greater influence over AI content generation. Some content producers are also striking content licencing deals directly with AI companies, of course.
