Are You Found in ChatGPT? Why Your Company Is Missing

A prospect wants to know who in the market combines operator consulting with AI implementation. They do not google anymore. They ask ChatGPT. The answer names three providers, one mechanic, one recommendation. Your company is not in it, even though that is exactly what you do. This is not a question about the future. This is Generative Engine Optimization, the discipline of being cited inside the AI answer itself, and it runs on different rules than classic SEO.

Before this sounds like hype: AI search is still small in 2026. In the German-speaking market, traffic from AI tools is only around 0.38 percent of total web traffic (SE Ranking, based on Similarweb clickstream across 101,000 sites). Anyone telling you ChatGPT has replaced Google is selling you something. The honest read is different. It is an early indicator, and it is moving fast.

Small, but the trend is unmistakable

Globally, referral traffic from AI tools more than tripled within a single year (Similarweb, September 2024 to September 2025). In parallel, 57 percent of German companies are engaging with AI, and around 44 percent of the population already use generative AI (Bitkom, bidt). So the users are there. What is missing sits on the supply side: content an AI can cleanly cite.

The reason this matters now, not in 2028, lies in click behaviour. In 2026, 68 percent of US Google searches end without a single click on a result (SparkToro, Rand Fishkin). Where Google serves its AI Overviews, organic clicks fell by 38 percent in a US field experiment (Search Engine Journal). Those are US figures, and the German-speaking market trails behind. But the direction is clear. The answer is moving from the link into the generated text. If you are not in the text, you no longer exist for the searcher.

And it is not enough to appear in ChatGPT alone. Within AI traffic, ChatGPT dominates the German-speaking market at around 65 percent, ahead of Gemini at roughly 16 and Perplexity at roughly 15 percent (SE Ranking, German-speaking market 2026). In B2B the picture shifts. A B2B panel from Goodie saw ChatGPT at 62.6 percent in 2026, but Claude already at 18.5 percent. Your buyers do not ask one machine. They ask four.

GEO is not SEO with a new name

SEO optimises for two things: a good ranking and a click. GEO optimises for a third: being cited, inside the answer the AI generates. That is not the same task with a fresh label. The rules partly run in opposite directions.

The proof comes from the cleanest study on this so far. Pranjal Aggarwal and colleagues at Princeton and Georgia Tech tested systematically in 2024, under the title "GEO: Generative Engine Optimization" (KDD 2024, arxiv.org/abs/2311.09735), which interventions raise a source's visibility in AI answers. Across thousands of queries, controlled, not from the gut. The results:

• Adding third-party quotations: up to 41 percent more visibility.
• Concrete statistics instead of vague claims: plus 33 percent.
• Clearer, more fluent language: plus 29 percent.
• Adding citations: plus 28 percent.
• Keyword stuffing: minus 8.4 percent.

Read that last line again. The old SEO move, packing keywords into the text, worsens your position in the AI answer. What a search engine used to reward, the generative engine punishes. That is exactly the proof that GEO is its own game.

The finding that matters most for smaller companies sits a little more hidden in the same study. For pages that rank low in classic search, simply citing sources delivered up to 115 percent more visibility. GEO helps smaller players disproportionately. You do not need to be the corporation with the heaviest domain. You need to be the one who backs up their claims.

Five levers you can apply this week

Flip the study around and it reads like a checklist. Five interventions, all proven, all doable without a tool budget.

1. Back every claim with a quote or a source

An AI builds its answer from passages that look like solid information. A claim without evidence reads as opinion, and it skips it. A statement with a named study, a quotation, an institution reads as citable. Plus 41 percent for quotations, plus 28 for sources, the single biggest lever in the whole set.

2. Write numbers, not adjectives

"Leading", "innovative", "market-leading" carries zero information for a machine. "0.38 percent of German-speaking web traffic, measured across 101,000 sites" is a fact it can pick up. Replace every value-laden adjective with the number behind it. Plus 33 percent visibility for concrete statistics.

3. Write clearly, briefly, without filler

Tangled marketing prose is hard for an AI to extract cleanly. Short sentences, one statement per sentence, clear terms. The study measures plus 29 percent for this. The same thing that keeps a human reader from bouncing makes the text machine-readable.

4. Name your sources visibly

Do not hide them in the footer; name them in the running text: who measured this, when, with which method. That gives the AI the anchor to tie your claim to a verifiable origin, and those are exactly the passages it prefers to cite.

5. Question as a heading, answer as a summary block

People ask AI tools in full sentences: "How do I find out whether my company shows up in ChatGPT?" A heading that mirrors that question, plus a short block that answers it directly in three or four sentences, is the unit a generative engine most likes to lift. That is why the TLDR block sits at the top of this page.

A word on two levers being sold everywhere right now. An llms.txt file, meant to tell AI crawlers what they may read, sounds tidy. An analysis across 300,000 domains found no measurable effect, and Google does not support the format. You can create one, it costs nothing, but do not sell it to anyone as a silver bullet. The same goes for schema markup. It is good practice for machine readability, but nobody can credibly prove a multiplier like "three times more citations". What is proven are the five levers above.

How MVA does it: this text is the proof

We do not claim GEO, we apply it to ourselves. The heading of this page is a question, not a slogan. At the top sits a TLDR block an AI can cite in three sentences. Inside the text are named studies, concrete percentages and dated sources, instead of "we are experts in AI visibility". The exact mechanic the article describes is built into the article. That is the difference between talking about GEO and doing GEO.

In the Web-Sprint we build pages structured to be citable from the start: question headings, summary blocks, clean schema, evidenced claims. In AI-Implementation we go a step further and set up monitoring that measures whether and where your company actually appears in ChatGPT, Perplexity, Gemini and Claude. The bigger reflex behind all this, that more volume solves structural problems, we cover in detail in Why more marketing won't help.

Ask yourself the honest question. If your next customer asks ChatGPT today who delivers the best in your market, is your name in the answer? If you do not know, that is the first thing you should measure.