What is GEO, Generative Engine Optimization

Generative Engine Optimization (GEO) is the set of activities that make a company, a brand or a person present, correctly cited and recommended in the answers of generative engines: ChatGPT, Gemini and Google's AI Overviews, Perplexity, Claude, Copilot. Where SEO works to get a page listed among the results of a search, GEO works to get the company's name inside the answer, with the right facts and the right sources.

The term was coined in 2023 by a Princeton University study (Aggarwal et al., GEO: Generative Engine Optimization) that measured for the first time how much certain features of a piece of content, such as statistics, source citations and quotations, increase the probability that a generative engine uses it in its answers. Since then the discipline has widened: it no longer concerns only the website, but everything the AI engines read about you.

Why AI answers have become a channel

Until yesterday, anyone who wanted to know something about a company opened Google and picked a link. Today a growing share of people ask an assistant directly and read the answer, without clicking. Some numbers, to be read as a trend:

• ChatGPT passed 900 million weekly users in February 2026, more than double a year earlier (OpenAI data reported by the press).
• Google's AI Overviews now appear on almost half of the searches tracked by industry studies, up more than 50% in a year.
• When an AI answer appears on the search page, clicks on results drop from 15% to 8% (Pew Research, 2025).
• Gartner estimates a 25% drop in traditional search volume in 2026 and up to a 50% reduction in organic traffic by 2028.

The consequence is simple: the AI answer is often the only thing your counterpart reads. A candidate, a buyer, an analyst, a journalist forms an opinion about you in thirty seconds, and that opinion was written by a model that read the sources it found, not the ones you would have wanted it to read.

GEO and SEO: what really changes

The two disciplines are not mutually exclusive: a technically sound website remains the base. But the data show that fewer than 10% of the sources cited by ChatGPT, Gemini and Copilot appear on Google's first page for the same question. Ranking first on Google does not guarantee being cited by AI, and vice versa. The detail is in GEO vs SEO.

How a generative engine works, in short

Every assistant builds its answers from three sources:

1. Training memory: what the model read when it was built. It includes Wikipedia, press, forums, company websites, but it is frozen at a date and weighs what was repeated across many sources.
2. Real-time browsing: when the question requires it, the model searches the web. ChatGPT and Copilot use the Bing index; Gemini uses Google; Perplexity has its own index. Here freshness, structure and readability of pages count.
3. Structured data and entities: business listings, Knowledge Graph, Wikidata, schema.org data. The model needs them to understand who you are and not to confuse you with others.

GEO works on all three sources, in a precise order: first measure what the AI engines answer today, then trace the sources that produce those answers, finally act on the sources. The method is described in How the GEO Assessment works.

What "being cited" means

In generative answers there are three levels of presence, worth distinguishing because they are measured and worked on differently:

• Mention: the company's name appears in the answer. It is the first goal for those who are absent today.
• Citation: the answer refers to a source (a link, a reference). Perplexity and ChatGPT with browsing almost always cite; Gemini shows the source links in AI Overviews.
• Recommendation: the AI proposes you as an option in a choice question ("which supplier", "where to apply", "whom to invest in"). It is the level that moves decisions.

To these add accuracy: being cited with a five-year-old news item, or with a wrong figure, can be worth less than not being cited at all. That is why the GEO Assessment measures presence, accuracy and sentiment together.

Who it matters most for

GEO is not just a marketing topic. Every audience asks the AI a different question about the same company:

• candidates ask what it is like to work for you (Employer Assessment);
• customers ask which product or service to choose (Brand Assessment);
• B2B buyers check whether you are a reliable supplier (Sales Assessment);
• journalists and stakeholders ask whether you are a serious, ethical company (Corporate Assessment);
• investors and analysts run due diligence (Investor Assessment);
• everyone asks who runs the company (Founder Assessment).

In life sciences, where ThatMorning has worked for over ten years, the regulatory perimeter imposes further rules: we cover them in the Life Sciences & Pharma section.

Where to start

By measuring. Before any intervention you need to know what the AI engines answer today, on which questions and from which sources. We do it with a free demo audit in 48 hours: real answers from several engines about your brand, an indicative GEO Score and the first sources identified. If you prefer to try on your own, the DIY test page has ten questions to paste into ChatGPT.