AI search news ·
Germany just ordered AI search to explain how it picks sources — the first legal lever on the citation black box
On 14 July 2026 the ZAK — the joint commission of Germany's 14 state media authorities — issued its first-ever administrative orders against AI services, naming Google's AI Overviews and Perplexity. The ruling: these are content providers, not neutral intermediaries, so German media law applies to them. ZAK chair Dr. Thorsten Schmiege put it in one sentence: “KI-Suchmaschinen und -Chatbots sind Inhalteanbieter – und wir wenden deutsches Medienrecht ab sofort konsequent auf sie an.” (“AI search engines and chatbots are content providers — and we now apply German media law to them consistently.”) The orders are not legally binding yet and Google has said it will appeal. We are covering it two weeks late and on purpose, because the part with consequences for AI visibility is not the headline everyone ran.
Primary source: die medienanstalten (ZAK) — official press release, 14 July 2026
What the orders actually demand
The headline coverage framed this as “AI answers are editorial content now.” True, and legally the load-bearing part — dropping the intermediary liability exemption is what makes everything else possible. But the obligations attached are more specific, and more interesting. Per the regulator's own release, the orders turn on transparency and non-discrimination in three concrete places:
- How links are made visible and findable — not whether sources exist somewhere, but whether a reader can actually get to them.
- Where journalistic content sits relative to the AI-generated answer — placement, not merely presence.
- How third-party sources are labelled in the link list.
These are the media-intermediary diversity rules of the Medienstaatsvertrag being pointed at an answer engine: state the criteria by which you select and present content, and do not disadvantage journalistic offerings without objective justification. A legal opinion by Prof. Dr. Jan Oster and Prof. Dr. Christoph Busch on AI integration in search engines was published alongside the decisions. Corroborated in English by Broadband TV News, 14 July 2026 and logged by Digital Policy Alert.
Why this is the interesting one
Every business trying to appear in AI answers is working against a black box. Nobody outside the vendors knows the selection criteria, which is why an entire industry — ours included — sells measurement of the output rather than knowledge of the mechanism. We infer it: 1,092 distinct domains across 511 ChatGPT answers, 96.8% of them appearing in only one vertical; 89.7% appearing in exactly one metro. That is archaeology, not documentation.
A transparency obligation attacks that directly. Not “label your AI content” — state the criteria by which you select and rank the sources you cite. No regulator has previously tried to compel that from an answer engine. If it survives appeal, the deliverable is a published account of how one of these systems chooses what to cite, which is the single document this entire field currently does not have.
What it means for you, honestly
- Nothing changes for your AI visibility today, and be suspicious of anyone selling you otherwise. The orders are not binding, Google is appealing, and German administrative proceedings are slow. No optimisation follows from this.
- The thing to watch is a document, not a ranking change. If a disclosure of selection criteria is ever published, it is the first primary source in this field — and it will immediately show how much of the prevailing GEO/AEO advice was inferred correctly and how much was invented. We would expect a fair amount of the second.
- Non-discrimination cuts toward publishers, not businesses. The protected class here is journalistic offerings. A dentist in Ohio is not a beneficiary. But the same disclosure that protects a newspaper describes the pipeline that decides whether the dentist gets named — which is why it is worth reading even from outside Germany.
Limits
- Not binding, and under appeal. Google has stated it will pursue legal remedies. Nothing here is settled law, and the orders may be narrowed or overturned.
- Germany only, and jurisdictionally narrow. The MStV binds these services in Germany. It creates no obligation in the US or UK, and any effect elsewhere would be a voluntary product decision by the vendor.
- We read the primary in German. Quotes above are from the ZAK release of 14 July 2026, with our translation shown alongside the original. Where we describe the three obligation areas, we are summarising the release's own wording, not the full text of the administrative orders, which are not public.
- ChatGPT is not named. The orders concern Google AI Overviews and Perplexity. OpenAI is not a party.
Until anybody is compelled to publish their selection criteria, the only way to know what an engine cites when a customer asks about your category is to ask it and read the answer. That is what the free 60-second check does; agencies covering several clients at once can use the $249 Agency 5-pack.
Will AI search engines have to disclose how they choose sources?
In Germany, possibly. On 14 July 2026 the ZAK ruled that Google AI Overviews and Perplexity are content providers subject to German media law, and issued orders requiring transparency and non-discrimination in how links are made visible, where journalistic content is placed relative to the AI answer, and how third-party sources are labelled. The orders are not yet binding and Google is appealing, so no disclosure exists today.
Does the German ruling change how I should optimise for AI search?
No. Nothing about it is in force, it applies only in Germany, and it creates obligations for the engines rather than opportunities for businesses. Any advice claiming a new optimisation tactic follows from this ruling is invented.
Does this apply to ChatGPT?
Not in these orders. The 14 July 2026 ZAK decisions name Google's AI Overviews and Perplexity's AI-powered search and news services. OpenAI is not a party, though the reasoning — that generating an answer makes you a content provider rather than a neutral intermediary — would not obviously stop at those two.
See your number
A free 60-second check shows what AI says about you.
Running this for clients? The $249 agency 5-pack audits five businesses, white-labeled.
Who runs this
- Built and operated by Sensara LLC, Atlanta, Georgia — about us and how the audit works.
- See what the report looks like before you run anything — score per engine, the competitors AI names instead of you, and a fix plan.
- We run the same audit on ourselves every week and publish the result: in the latest run AI named AskedAbout in 1 of 144 answers. We report our own numbers the way we report yours.