The Ruling
On July 14, 2026, Germany's Kommission für Zulassung und Aufsicht (ZAK) — the Commission for Licensing and Supervision that oversees the country's regional media authorities — issued the first regulatory orders anywhere classifying AI-generated search summaries as the provider's "own content" rather than passively hosted material (heise.de). The proceedings, brought jointly by the Hamburg/Schleswig-Holstein and Berlin-Brandenburg media authorities, targeted Google's AI Overviews and Perplexity's AI-powered search and news product. ZAK Chair Thorsten Schmiege put the finding bluntly: "AI search engines and chatbots are content providers, and we are now consistently applying German media law to them."
The practical consequence is significant. Because the AI summaries are now treated as editorial output rather than intermediated third-party content, the liability exemptions the EU's Digital Services Act extends to hosting and caching services no longer apply, and both companies instead fall under the Medienstaatsvertrag's accuracy, transparency and non-discrimination provisions for outlets with "particular relevance for the formation of public opinion" (PYMNTS). Google has said it will appeal; Perplexity, which reportedly lacks a designated legal representative in Germany, declined substantive comment. Both companies have one month to file legal remedies against the orders.
The Case for the Ruling
The regulators' concern is not manufactured. AI Overviews physically displaces the "ten blue links" that once sat atop a results page, replacing them with a synthesized paragraph that answers the query directly — often without the user ever clicking through to the outlet whose reporting supplied the answer. If that summary misattributes a claim, buries a competing viewpoint, or simply starves publishers of referral traffic, the DSA's Article 6 hosting exemption — built for platforms that store or transmit someone else's content passively — is a poor fit, because Google and Perplexity are actively authoring the text a user reads. Legal scholars Jan Oster and Christoph Busch have argued the shift from linked results to single prose answers is a structural threat to journalism's funding model, not a cosmetic one, and that AI search may warrant its own regulatory category rather than inheriting rules built for search engines or social platforms (the-decoder.com). Germany's media-plurality tradition — rooted in decades of broadcasting jurisprudence protecting Meinungsvielfalt — gives regulators a genuine, non-pretextual reason to worry when one company's product design decides which outlets get seen.
Why Media Law Is the Wrong Instrument
The steelman only goes so far. Reclassifying an entire AI product category as "content" under a framework designed for licensed broadcasters and newspapers creates exactly the fragmentation the DSA was built to prevent — a single EU-wide liability regime enforced by a home-member-state authority, rather than sixteen German Länder-level media authorities, let alone 27 EU states, each independently deciding whether a chatbot's output is "editorial." Perplexity's absence of a German legal representative is itself evidence of the enforcement gap: a media-law order against a company with no local presence risks becoming symbolic rather than operative, unless Germany persuades other jurisdictions to follow suit.
There's also a perverse-incentive problem. If AI summaries trigger broadcaster-grade accuracy and plurality obligations the moment they cite or link to a publisher, the rational response is to cite and link less — stripping attribution rather than adding it, the opposite of what media-plurality advocates want.
"Anyone who controls whether content gets found through the selection and placement of links must make that transparent. Otherwise, diversity among journalistic and editorial outlets will disappear." — ZAK Chair Thorsten Schmiege
The Better Precedent Was Already Sitting in Bonn
Germany doesn't need a novel media-law theory to discipline self-preferencing by dominant platforms — it has one already working. Since 2021, Section 19a of the GWB has let the Bundeskartellamt designate companies of "paramount significance for competition across markets" and then prohibit specific abusive conduct, including self-preferencing, at an earlier stage than ordinary antitrust law allows. Five companies — Alphabet/Google, Meta, Amazon, Apple and Microsoft — carry that designation today, and the Federal Court of Justice has upheld the underlying framework against both Amazon and Apple (Bundeskartellamt). In February 2026, the authority used that exact provision to fine Amazon €59 million and order disgorgement of profits from its Marketplace price-control mechanisms (Jones Day) — a targeted, judicially reviewable remedy aimed at specific conduct, not a wholesale reclassification of a product category.
If AI Overviews' placement genuinely amounts to self-preferencing that suppresses journalistic discoverability, that is a Section 19a case, or a DMA Article 6(5) self-preferencing case against Alphabet as a designated gatekeeper — both narrower, both already tested in German and EU courts, and both avoiding the collision with DSA harmonization that the ZAK ruling now invites. Regulators concerned about journalism's survival should reach for the tool that has already produced enforceable, appeal-tested outcomes, rather than one that starts its life with both defendants heading straight to appeal.