Ten mistakes when choosing a GEO course

By Miguel Maio · Published 19 October 2026

Selling a guaranteed spot in AI answers is selling what no one controls. Google states plainly that it does not accept payment to rank you higher, and that anyone who tells you otherwise is wrong; and the models generate answers probabilistically, with no spot for sale. It was the promise my students most often arrived with when I built the GEO course at Lisbon Digital School, in June 2026, one of the first in Portugal. Much of the class went into dismantling it, and the mistakes that travel with it.

The ruler, before the ten mistakes

An honest GEO course teaches three things you can verify: building your entity, making your information citable, and measuring what happens. Everything else - hacks, guarantees, ChatGPT secrets - either can't be proven or can't be controlled. Keep the ruler: each of the ten mistakes is paying for something that doesn't pass it.

The ten mistakes

Buying what no one controls

1. It promises guarantees. "I guarantee a ChatGPT citation." "First place in AI Overviews." Run. An LLM's output is probabilistic and shifts from one day to the next, and Google itself says there is no placement for sale. The honest maximum is raising the probability, and telling you that's all it is.

2. It mistakes being everywhere for being cited. It sells ubiquity when 58.5% of Google searches don't even produce a click (SparkToro, 2024). What counts is being the right answer to the right questions, not filling engines.

3. It sells llms.txt as a silver bullet. It's an emerging proposal, with no guarantee any engine uses it. A good course treats it as a curation discipline. A bad one promises it saves you.

Not understanding the terrain

4. It treats "AI" as one block. AI Overviews, AI Mode, ChatGPT and Perplexity cite different sources: between AI Mode and AI Overviews the URL overlap is 10.7% (SE Ranking, 2025). Teaching you to optimise for AI as a single target is teaching half a dozen games as if they were one.

5. It confuses AEO with GEO. Optimising for Google's AI features is one slice. GEO spans every engine. Swap the names and you swap the strategy.

6. It skips the entity. GEO rests on who Google and the models recognise: schema, consistent sameAs, coherent presence. Going straight to tactics without this is building on sand.

No measurement, no proof, no honesty

7. It teaches you to appear, not to measure. Without AI channels in GA4 and citation tracking, you don't know if it worked. Without measurement, GEO is faith.

8. The trainer isn't cited by AI. Whoever teaches you to be cited by the models should show up in them. Ask ChatGPT or Perplexity about them; search the Knowledge Graph and search results. Silence is an answer.

9. Museum material. In a field that changes monthly, a programme with no visible update date is a risk. Ask for the date. They won't give it? You already know.

10. It never tells you when it's not worth it. The course with no verdict, that never admits what it can't control, is selling you comfort.

The unbreakable limit: method and measurement

No course gives you the certainty of a citation. What a good one gives you is a repeatable method and a way to measure. Mine is CITAR: Clarity of entity, Information that's citable, Trust (E-E-A-T), All-engine reach and Results tracking. None of the five letters is a trick. They're groundwork you can verify, front by front, with measurement at the end.

What you lose if you choose wrong

Choose wrong and you don't just lose the course fee. You lose a year applying tactics no one measures, while whoever learned to measure is already fixing what you can't even see. In the classroom, in June, this is what I took to the students: method and measurement, not tricks. A course that guarantees you the result is selling what isn't theirs to sell.

Before you pay, run this list

• 01

  Guarantees

  Promises a guaranteed citation or position? Close the tab.

• 02

  Ubiquity

  Talks about being everywhere, or winning the right questions?

• 03

  llms.txt

  Sold as a certainty, or as curation with no guarantees?

• 04

  "AI" as one block

  Separates AIO, AI Mode, ChatGPT and Perplexity, or lumps them together?

• 05

  AEO vs GEO

  Distinguishes the two, or uses them as synonyms?

• 06

  Entity

  Covers schema, sameAs and E-E-A-T, or jumps straight to tactics?

• 07

  Measurement

  Teaches GA4 for AI and citation tracking, or just appearing?

• 08

  Trainer

  Shows up in AI answers and the Knowledge Graph?

• 09

  Update

  Does the programme have a recent, visible date?

• 10

  Verdict

  Tells you when it's not worth it, or only sells?

Sources

• SparkToro / Search Engine Land - 2024 Zero-Click Study
• SE Ranking - AI Mode research (2025)
• Google Search Central - How Search Works

Frequently asked questions