64.9% of German AI Search citations point to a domain that carries a local region or language signal: a .de, .at or .ch address, or a global domain on a German URL such as .com/de-de/. That is from the OtterlyAI German AI Search Index, a study of 16 industries and more than 3.95 million citations across seven AI search engines.
The more useful question is which German sites those citations go to, and the answer does not look like the American web. The media publishers that get cited are comparison and test titles, Finanztip, Chip, Test.de, Computerbild, rather than general news. Reddit, which carries 1.79% of American citations, carries 0.60% for German ones. Associations take 8.9% of German citations against 1.8% for government sites: Verbraucherzentrale, ADAC, the IHK network, vfa.de.
That last contrast is why US numbers run through this piece. Most of the GEO advice reaching German desks was written from American data, and this is where it stops applying.
How German AI Search behavior is different from the US, in short:
- German news reading is consumer-test shaped. Finanztip, Chip, Handelsblatt and Test.de lead the German news list. The American list is business media: Forbes, CNBC, Bankrate, NerdWallet.
- Reddit is an English channel. 0.60% of German citations against 1.79% of US ones, three times the weight over there.
- YouTube carries over at parity. 1.99% in Germany against 1.75% in the US.
- Company-owned pages are the majority in both markets, 57.2% here and 56.3% there. The category mix barely moves. What moves is the nationality of the source.
- German prompts cite associations where American prompts cite government. Associations take 8.9% of German citations against 5.3% in the US; government sites take 1.8% here against 3.1% there.
- 64.9% of German citations carry a language or region signal. 51.1% are local domains on .de, .at or .ch, against 0.3% on .us in the United States.
- A global competitor can get through with a localized URL. 13.8% of German citations go to a /de/ path or a de. subdomain on a global domain. That is one localized URL for every 3.7 citations on a local domain.
- Industry decides more than the country does, inside Germany. Healthcare sends 73% of its German citations to local sites, Educational Platforms 31%.
- Every engine prefers German sources, but not equally. The share of an engine’s German citations landing on a German source runs from 75.2% at Perplexity to 57.7% at Claude.
Where German AI citations actually come from (vs the US)

1. Company owned content
Company-owned content is the majority of what gets cited, 57.2%, and that is true in the United States too at 56.3%. The mix is not where the markets differ. What differs sits underneath it.
2. News and media
The media publishers that get cited are the ones that compare things. News is marginally lower in Germany at 14.9% against 16.4%, but the composition is the real story.

The German list is dominated by consumer-test and comparison publishers: Finanztip takes 4.2% of all German news citations, Chip 3.4%, Handelsblatt 3.3%, Test.de 2.9%. The American list is business and technology media: Forbes at 5.3%, CNBC at 2.5%, Bankrate at 2.2%, NerdWallet at 2.1%.
3. NGO, associations and governmental sites
NGO and association sources take 8.9% of German citations against 5.3% of American ones, the largest single gap in the mix.

That 8.9% is not an abstraction, and it is concentrated. Verbraucherzentrale alone takes 3.6% of every German association citation, ADAC 2.8%, then vfa.de at 1.3%, KfW at 0.9%, the IHK network at 0.8% and the Paul Ehrlich Institute at 0.8%. Consumer advice centres, the automobile club, an industry body, a development bank, chambers of commerce, a regulator. The American equivalent is thinner and more commercial, led by consumerreports.org at 5.1% of US association citations.
Government sites run the other way: 3.1% of US citations against 1.8% of German ones, carried by fda.gov, sec.gov and the NIH library. Combine the two categories and Germany still comes out ahead, 10.6% against 8.4%, but it routes that institutional trust through a different kind of body.
4. Social media, forums, video and community
Forums are not your channel here.

Community and forum sources take 1.5% of German citations against 2.8% of American ones, and Reddit alone is almost the entire difference: 0.60% against 1.79%. Gutefrage, the closest German equivalent, reaches 0.08%. YouTube does carry over, at 1.99% against 1.75%, so video advice written from US data still applies and forum advice does not.
5. Education and reference run slightly German
Education platforms take 3.8% of German citations against 2.8% of American ones, and reference and encyclopedia sources 1.8% against 1.3%. Academic sources run the other way, 0.4% against 0.9%, because PubMed Central and the American journal web have no German counterpart at that scale in these answer sets.
The most cited German domains (vs the US)

Nine of the fourteen most cited domains in Germany take no measurable share of the American market at all. The two markets share a platform layer, YouTube, Google and Wikipedia, and almost nothing else. Reddit is the clearest split: cited three times more often for English US prompts than for German ones.
64.9% of German citations carry a language or region signal

For German prompts, every citation lands in one of four places:
- 51.1% on a local country domain, .de, .at or .ch
- 13.8% on a global domain carrying a German signal in its URL, 11.9% on a path such as /de/ or /de-de/ and 1.9% on a subdomain such as de.statista.com
- 34.1% on a global domain with no language signal at all
- 1.0% on another country’s domain
The first two lines are the 64.9%. The other 35.1% is the part worth sitting with: more than a third of everything AI Search cites for a German prompt carries nothing in the address to say it is German, and a global page competes for it on equal terms.
Strip out the platforms that never localize by path, YouTube, Google, Reddit, Wikipedia, Facebook, Instagram, LinkedIn, and the local share rises to 54.0% while the localized share holds at 14.0%. The pattern is not an artefact of platform URLs.
If you run a .de, the first line is the good news and the second is the one to watch. Local domains still outnumber localized URLs 3.7 to 1, so the home advantage is real. But the international competitor entering your market does not need to buy a German domain to appear next to you. It needs a localized path on a domain that already has authority.

Read that chart one row at a time. Each bar is the whole of what AI Search cites from that kind of site in Germany, so the pink and purple together are the German-signalled share and the amber is what is left for a page with no German address at all.
Two rows matter most. Associations are 77% local, the most German corner of the market, and almost nothing reaches them from a localized URL. Company and brand pages are 57.2% of everything cited, and they split 48% local, 19% localized, 32% with no signal. That 19% is the open door: of the German citations company pages send to global domains, 37.8% land on a German URL, against 28.9% across the market as a whole.
Individual sites show what that looks like. Qonto takes essentially all of its German citations on localized URLs, apps.apple.com 85%, fortunebusinessinsights.com 74%, mordorintelligence.com 60%, coursera.org 43%. Statista does it with a subdomain and takes 0.52% of all German citations that way, the fifth most cited domain in the market.
The American control makes the point. Of American citations to global domains, 5.1% sit on a localized URL. A German path is not what URLs look like everywhere. It is what they look like when a site is competing for German answers.
Some industries point more to German sites

Healthcare sends 73% of its German citations to local domains, Finance and Insurance 69%, Retail and E-commerce 66%, Consumer Goods 66%. At the other end, Educational Platforms sends 31%, Industrial and Manufacturing 37%, Transportation 39% and Hospitality 39%.
Read your own row before you read the market average. At 73% local, a German healthcare site is competing almost entirely against other German sites and a global entrant has little room. At 31%, a German e-learning site is competing against the global web, and being legible to an engine reading mostly English sources is part of the job.
Every AI search engine prefers German sources, but not equally

The chart counts, for each engine, the share of its German citations landing on a German source: a .de, .at or .ch domain, or a global domain on a German URL. No engine falls below 57.7%, and the spread is 17.5 points, from 75.2% at Perplexity down to 57.7% at Claude. Run the same measure on American prompts and it is 3.6% to 9.4%, where the engine barely matters.
Underneath that single number the engines differ again. On local domains alone, Perplexity sends 55.9% of its German citations there and ChatGPT 45.2%, the widest gap in the set. ChatGPT also sends 39.3% of its German citations to global domains with no language signal at all, against Perplexity’s 23.8%. A global English page has its best chance in ChatGPT and Claude and its worst in Perplexity.
Two more differences sit underneath the geography.
Concentration: half of all American citations in Google AI Mode come from 301 domains, and the same engine in Germany needs 588. Gemini shows the same pattern, 508 against 738. The German incumbent set is less entrenched, which is an opening.
Share of voice: Gemini is the second largest source of citations in Germany at 17.5% but fourth in the US at 16.3%, while Google AI Mode runs 9.6% here against 10.8% there.
Conclusion: the GEO playbook for the German region
If you are an international company expanding into Germany
Build /de/ before you buy .de. 13.8% of all German citations already go to localized URLs on global domains, and for company pages that route carries 37.8% of their global-domain citations. A localized path on a domain that already has authority starts from a stronger position than a new country domain with none. If you already own the .de, run both; the point is sequencing, not exclusivity.
Localize the URL, not only the page. The classifier in this study reads the address. So, in effect, do the engines: the correlation between a German-language path and a German citation is the strongest structural pattern in the data. A German page served at the same English URL is invisible to that signal.
Translate the pages that answer buying questions, not the brochure. The prompts behind this data are bottom-funnel. Comparison pages, pricing, specifications and support documentation are what gets cited.
Do not benchmark Germany in ChatGPT alone. It is the least local engine on domains, 45.2% against Perplexity’s 55.9%. If your German reporting runs on one engine you will systematically understate how much local presence you need. Measure all seven, or at minimum pair the most local engine with the least.
Get into the association layer. Associations carry 8.9% of German citations against 5.3% in the US, and they are the category an international entrant is least likely to have on its plan. Industry bodies, chambers of commerce, sector associations. Membership, published data, and being quoted in their material are all routes.
Set expectations by industry, not by market. If you are in healthcare or financial services, around 70% of what the engines cite is German and a global estate will not carry you. If you are in e-learning or industrial manufacturing, a well localized global domain competes on near-equal terms.
Reddit is not your German channel. It carries a third of the citation weight it carries in the US. YouTube does carry over, at parity.
If you are a German business serving German customers
Your .de is an asset, so use all of it. You start inside the 51.1% of citations that global competitors have to fight their way into. The gap is usually depth of content, not domain authority.
Depth beats breadth, because the German tail is long. It takes 588 domains to make up half of Google AI Mode’s German citations against 301 in the US. The incumbent set is less entrenched, which means a thorough page on a narrow question has a real chance.
Answer the comparison question directly. The sites winning German citations are overwhelmingly comparison and test publishers: Finanztip, Chip, Test.de, Computerbild, Check24, Verivox, Vergleich.org. If you sell a product, the page that compares it against the alternatives is the page that gets cited.
Pursue association and consumer-test visibility. Verbraucherzentrale, ADAC, the IHK network, sector associations like vfa.de. In the German answer set these function the way government and major media do in the American one.
Do not copy American GEO advice on forums. Advice to build Reddit presence is written from a market where Reddit carries three times the weight. Put that effort into YouTube, which performs in Germany at the same level as in the US, and into the test and comparison publishers.
Watch the 35.1%, not only your 51.1%. More than a third of German citations go to addresses with no German signal, and that is the share a well localized international competitor is aiming at. Your industry row tells you how exposed you are: at 73% local a German healthcare site has room, at 31% a German e-learning site is already competing against the global web.
About the AI Search Index: How this was measured
Read the full study here: OtterlyAI German AI Search Index
– Window: 16 August to 28 September 2026, 44 days, identical dates in both markets.
– Scope: the 16 industries tracked in both Germany and the United States. The US also runs a 17th, AI Data Governance, which is excluded so the comparison is 16 against 16.
– URL level, not domain level. 7.7 million citations, 3.95 million German and 3.73 million American. The domain endpoint caps at 400 domains per report, engine and day. The URL endpoint has no cap, and it is the only way to see a path.
– Seven engines in both markets: ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity, Microsoft Copilot and Claude. Prompts are bottom-funnel commercial questions, around one hundred per industry, of the kind someone asks immediately before buying.
– How a domain is classified. A .de, .at or .ch address is a local country domain. Everything on a global domain is then read for a language signal in the first part of the host name or the first three segments of the path, matched whole, so /design/ is not German and deliveroo.com is not either. .de accounts for 97.4% of the local bucket; .at and .ch together are 2.6%.
– The American figures are a control, not a benchmark. They are here to show which German patterns are specifically German. 28.9% of German citations to global domains carry a German URL, against 5.1% of American citations carrying an American one.
Got a GEO experiment idea? Let’s collaborate
At OtterlyAI, we don’t write about AI Search from the sidelines. We test it. With the community, we run and update a public GEO Experimentation Tracker showing what we are running and what moved AI Search. If you are running your own GEO experiment, email rick.tousseyn@otterly.ai and we may feature it with your name attributed.
Want to measure your AI Visibility? Start with OtterlyAI and our GEO Guide.
Learn how to connect your Claude or ChatGPT to OtterlyAI’s MCP.





