GEO AI in 2026: what it is, how it works and how to get into ChatGPT, Gemini and Perplexity answers

Picture this: a mom is looking for an online speech therapist for her child. She used to open Google and scroll through ten websites. Now she types into ChatGPT: “recommend an online speech therapist for a 5-year-old, we live in Poland” and gets three names. If your school isn't one of them, the customer will never even hear of you.

At MAS Project we were among the first in Ukraine to apply GEO to client businesses. In this guide we explain how AI recommendations work and what exactly you need to do.

What GEO is and how it differs from SEO

The term Generative Engine Optimization appeared in 2023 in a study by researchers from Princeton and other universities. Generative systems (ChatGPT, Gemini, Perplexity, Google's AI Overviews) don't give a list of links; they write a ready answer. GEO is about optimizing so your brand makes it into that answer.

Why you should start GEO now

• The audience is huge. According to OpenAI, in early 2026 about 900 million people used ChatGPT every week.
• Clicks are leaving search. A Pew Research Center study found that when Google shows an AI summary, people click a regular link on 8% of visits, compared with 15% without one.
• AI traffic is still small but concentrated. According to Statcounter, in August 2026 AI chatbots accounted for about 0.25% of all website referrals worldwide, and almost 80% of that came from ChatGPT. That means the spots in AI answers aren't taken yet, and competition is lower than on Google.
• Whoever gets in first gets remembered. AI relies on accumulated data. Brands that build quality mentions of themselves now gain an advantage for years.

How AI assistants choose whom to recommend

An AI answer comes from two sources:

1. Knowledge from training. What the model “read” on the internet up to a certain date: websites, articles, directories, forums. It's updated with new model releases.
2. Search at the moment of the question. ChatGPT, Gemini and Perplexity look up fresh pages online and summarize them. You can get in here much faster.

In both cases the AI trusts brands that:

• have clear facts: what the company does, for whom, where and with what results;
• have consistent information across sources: one name, one description, the same contacts on the website, in directories, on social media and on maps;
• others write about: directories, review platforms, media, expert articles, discussions;
• have reviews: real customers on Google Maps and industry platforms.

If there's little data about a company or it contradicts itself (different names, different founding years, different figures), the AI prefers not to take the risk and recommends a competitor.

What the Princeton study found

The authors of “GEO: Generative Engine Optimization” (KDD 2024) tested which text changes increase the chance that a generative system cites a source. Three techniques worked best: adding citations to sources, adding quotes from experts and adding statistics. According to their data, these methods increased a source's visibility in answers by up to 40%. Classic keyword stuffing, on the other hand, barely helped.

The practical takeaway: AI likes texts that look like a reliable source, with specific numbers, facts and links rather than generalities.

7 steps of GEO

1. Visibility audit

We put together 20–50 questions your customers ask and check the answers from ChatGPT, Gemini, Perplexity and Google AI Overviews. We look at whether you're mentioned, how you're described, who's recommended instead of you and which sources are cited.

2. A single brand “passport”

One spelling of the name, one description, the same contacts, founding year, team and numbers across your site, Google Business Profile, LinkedIn, directories and social media. For an AI, that's a signal it's dealing with one real company.

3. A site machines can read easily

Pages built around specific customer questions, FAQ blocks, Schema.org markup (organization, services, reviews, articles), an llms.txt file with a short company description for AI systems and a clear heading structure. Key facts go in the text, not only in images.

4. Content with facts and sources

Expert articles with numbers, case studies and links to research. According to the Princeton study, this format is more likely to make it into answers.

5. Mentions on third-party platforms

Agency and company directories, industry rankings, review platforms, media columns, podcasts, YouTube videos. The more authoritative places describe you consistently, the more confidently the AI recommends you.

6. Reviews

Regular, genuine reviews on Google Maps and industry platforms. AI uses them both as proof that the company exists and as an argument in its recommendation.

7. Monthly monitoring

Every month we re-check the same questions and track the trend: whether you've appeared in answers, how the description changed and which sources are cited.

How long until results

Who needs GEO most

• Education: schools, courses, tutors. Parents actively ask AI where to send their child.
• High-ticket services: clinics, lawyers, agencies, B2B. People want advice before an expensive decision.
• Experts and online course creators: when it matters that the AI knows the author and their method.
• Companies working across several countries: AI answers in many languages, so you can build visibility in several markets at once.

GEO mistakes

• Mass-generated texts without facts. AI doesn't cite empty content.
• Fake reviews and mentions. A risk of bans on platforms and lost trust.
• Contradictory company data. Different numbers on your site and in directories are the biggest enemy of recommendations.
• Expecting results in a week. GEO builds up over time, just like SEO.

GEO and ChatGPT Ads

These are two different tools. GEO is the free recommendations the AI gives on its own. ChatGPT Ads are paid, labeled ads next to the answer that OpenAI launched in 2026. They work best together: ads bring traffic right away, GEO a steady stream of recommendations over the long term.

Sources

1. Aggarwal et al.: GEO: Generative Engine Optimization (Princeton, KDD 2024)
2. Gartner: search engine volume will drop 25% by 2026 due to AI chatbots
3. Pew Research Center: Google users are less likely to click links when an AI summary appears (2025)
4. Statcounter: AI chatbot share of website referrals
5. DemandSage: ChatGPT user statistics 2026