Which are the GEO agencies in Berlin?

Which GEO agencies in Berlin do the engines name?

New Map's measurement counted 7 answers to "GEO agencies in Berlin" across Perplexity, Gemini and ChatGPT, over 2 weekly readings. That is one week's reading of 68 questions about AI visibility tools on 3 engines (146 answers), not a trend: Perplexity cited a source in only 7 of its 68 answers, and Claude Code's 10 answers are too few to read alone. The engines named none of the tracked companies, and New Map was named in 0 of the 7. Method: how New Map measures.

What the engines gave was sources. The table lists the domains cited, as the measurement reports them.

These domains are unverified and not endorsed. A citation says nothing about what a site sells or how well it works. Use these criteria instead:

• It shows its method: questions, engines, runs.
• It shows the sources engines cite for its clients.
• It shows what engines say when asked about it by name.

The chart shows how often each domain was cited among the 7 answers.

Do the engines give the same Berlin agency answer twice?

Not reliably. New Map's measurement split the 7 answers as Perplexity 2, Gemini 3 and ChatGPT 2, so no engine has enough answers to read alone.

In one run versus the next, AirOps² (undated) found that brand visibility in AI answers held steady for only 30%. When the same prompt was executed five times in a row, AirOps found that 20% of brands stayed visible across all five runs (undated).

SparkToro's January 2026 test of 2,961 runs found ChatGPT and Google AI repeated the same brand list in under 1 in 100 runs.

So an agency that says an engine "recommends" it after one prompt has shown you one draw. Ask how many runs, on which engines and over how many weeks its claim rests.

Which sources do German AI answers cite for a Berlin GEO agency?

Otterly³ analysed 3.95M citations across 16 industries and seven engines in October 2026. German-signalled domains (.de, .at, .ch or a German URL signal) took 64.9% of German AI citations. Another 34.1% went to global domains with no signal.

Engines differ. Otterly.ai measured how Perplexity routed its German citations: 75.2% went to sources marked as German (same study). The chart sets that engine against the all-engine figure.

This fits what New Map saw: geoagenturberlin.de, happyheroes.de, suxeedo.de and geoagenturen.de all end in .de, and all were cited for the Berlin question.

Company-owned pages made up 57.2% of German citations in the same Otterly study. Our read is limited to this: the agencies' own German pages are a large part of what engines draw on in Germany.

How do you tell one Berlin GEO agency from another?

Chen and Koudas, a University of Toronto preprint, measured 158,099 answers. Where a page came back in ChatGPT's own search explained 32% of which pages from one site it cited, while all page features combined explained 2.8% (Chen, October 2026).

That study measured page features in one preprint, and its authors advise a GEO vendor. Ask what an agency changes beyond formatting, such as which sources it earns and which questions it tracks.

Platforms are also moving. DEJAN AI⁴ reported that DeepSeek's search engine launched on 17 October 2026 and that its API documentation lists GEO alongside SEO spam as content its index filters out. That is a platform statement, not a test of what works.

The diagram sets out a vetting order. The first step rests on how a brand's visibility in AI answers is measured.

Where does a Berlin brand need mentions for engines to name it?

The study is not about agencies. Applied to agency vetting, it is an inference: what others publish about an agency may matter more to the engines than what the agency says about itself.

Muck Rack's May 2026 analysis of more than 25M cited links on ChatGPT, Claude and Gemini put the earned share of citations on unbranded questions at about 84%. "Earned" there means every source that is not owned or paid, including Wikipedia and social sites, and Muck Rack warns it does not mean PR drives that figure.

So ask a shortlisted agency to show press, community and directory placements outside its own site, and which of them the engines cite. An agency with only its own pages to show has shown you one source type.

What to do

1. Run "GEO agencies in Berlin" yourself on Perplexity, Gemini and ChatGPT several times, and record which domains each engine cites, not only which names appear.
2. Ask each shortlisted agency for its measurement: the questions, engines, number of runs and weeks behind any visibility claim, as AirOps found only 20% of brands stay visible across five runs.
3. See plans and pricing if you want New Map to measure what engines answer about your own brand.

Frequently asked

Sources

1. New Map measurement, How we measure, 19 October 2026
2. AirOps, The GEO Strategy Playbook for Marketers and Organic Growth Teams, airops.com
3. Otterly, AI Search in Germany Plays by Different Rules. Here Is the Data., otterly.ai, 5 October 2026
4. DEJAN AI (Dan Petrovic), DeepSeek Search Engine, dejan.ai, 18 October 2026
5. Buecheler et al. (arXiv), A Cue's Hidden Price: Auditing Seller Influence on AI Shoppers, arxiv.org, 19 October 2026
6. Chen et al. (arXiv), What Drives AI Citations? Content, Retrieval and Model Knowledge in Generative Engine Optimization (GEO), arxiv.org, 10 October 2026