Responsible AI Search? Google AI and an interesting ethical question

Taking a train in Switzerland, I sometimes read the 20min news magazine that is freely available at most train stations (the idea of the name is that it takes you a 20-minute train ride to read, which is actually quite accurate, I find).

Anyway, I came across an interesting article titled “Googles KI-Modus: Medienverband warnt vor dem neuen Modell” (English: “Google AI-mode: Media organizations warn of the new model”, here is the online article for reference).

Apparently, media organizations are complaining that they are losing many clicks and site visits due to Google’s recent AI summarization mode. The new feature not only searches for relevant sources anymore but now also presents a short summary of the retrieved data. Probably from at least the first ten results or so.

While the sources of these AI summaries are still linked, it seems that many people don’t bother to click on them and look at the original sources directly. Perhaps the summaries are just that good and immediately fulfill the user’s information need. Also maybe we humans simply tend to be a little lazy at times and don’t bother to look further. Either way, the outcome is that now the waiter gets all the credit for the food: Google retrieves, summarizes, and serves, while the actual content cooks are starting to miss user engagement and, let’s be honest, also ad revenue.

The Cook and the Waiter

I was wondering whether I have any stance on this, and I can actually see arguments for both sides. The side of the content provider is quite clear. But the perspective of Google can also be understood.

If Google stops using AI summaries, people may stop using Google, as many are already used to such summaries — after all, ChatGPT can also search and summarize. There is competitive pressure, and the waiter wants to stay relevant in the restaurant business.

The ethical question for me is hidden in this cook–waiter analogy. If the food is excellent, we would probably like to attribute the cook more. If the service is excellent, then we also like to attribute the waiter, which, in many countries, is customary in the form of tips anyway.

In this analogy, Google is the waiter, whereas the content providers are the cooks. Now, how can the cooks be attributed if their role is more and more hidden? What do users even want to have on their “digital menu”?

Some may merely want to get a quick overview, and they decide to trust summaries — which are ideal for this — more and more. Others would still like to retrieve the original sources. What’s the ratio of these user needs? 80/20 maybe? These questions seem intermingled and may strongly influence how a good attribution balance between cook and waiter should be established.

When the Waiter Starts Cooking

Also, Google kind of becomes a little cook on its own when it merges multiple sources together into something that is novel (a summary). The line between serving and cooking gets blurred here.

If the AI system starts to create something genuinely new, even just in the structure or combination of facts, is it still only the waiter. Or is the AI already a sous-chef in the kitchen?

This, for me, is the real ethical inflection point. Summarization is not a neutral act anymore, especially if it’s multi-document; it’s a kind of re-creation of knowledge.

Responsibility and Credit Attribution

For me, a challenge of responsible AI use is reflected exactly in this balance. On the one hand, users want convenience — they want their dish served fast, warm, and nicely plated. On the other hand, if nobody credits or pays the cooks anymore, what then?

So maybe the solution is not about choosing between waiter or cook but about finding a way for both to thrive. Perhaps we need visible attribution, fair sharing models, or user options to switch between “summary view” and “source view.”

In other words: if we only tip the waiter, we might soon end up in a restaurant with no kitchen.

---

PS: I’m sure that the restaurant analogy maybe isn’t sound in every way, but I thought it’s a reasonably good fit.