Capstone · Beginner
Capstone: Break into GEO
The Break into GEO capstone is a hands-on AI visibility audit: you choose a brand, test 20 prompts across 3 AI engines against 3 competitors, then write a one-page findings memo and rewrite one page using answer-first principles. It brings together everything from the three courses. Plan for 6–10 hours (estimate). You'll be scored on a published rubric; 70% passes.
What's the brief?
- Choose a brand. Any company with a website and real competitors. If you use your employer's brand, get permission to share the results.
- Pick 3 competitors. Use the brands buyers actually compare it with, not the ones the brand wishes it competed with.
- Build a 20-prompt set. Mix the stages from GEO 101: category questions, comparisons, problem-led questions and brand questions. Write prompts the way a real buyer would.
- Run the prompts across 3 engines. For example ChatGPT, Perplexity and Google AI Overviews. Use a MentionX free account or a manual sheet. Record whether each brand is mentioned, cited or recommended, and which sources are cited.
- Write a one-page findings memo. Cover where the brand appears, where competitors win, which sources drive answers and three prioritised actions.
- Rewrite one page. Choose the page most relevant to a prompt where the brand loses. Apply answer-first writing, question-style headings and clear facts. Submit before and after versions.
What must you submit?
| Item | Format |
|---|---|
| Prompt set and results | MentionX export or spreadsheet: prompt, engine, date, mention/citation/recommendation per brand, cited URLs |
| Findings memo | One page, PDF or doc |
| Rewritten page | Before and after, with a short note on what changed and why |
How is it scored?
| Criterion | Points | What earns full marks |
|---|---|---|
| Prompt set quality | 20 | Realistic buyer phrasing, all four prompt types, no leading prompts |
| Data accuracy | 20 | Consistent recording, dates and engines logged, cited URLs captured |
| Analysis | 25 | Clear patterns, competitor gaps explained, source types identified, limits of a 20-prompt sample stated |
| Recommendations | 15 | Three specific, prioritised actions tied to findings |
| Page rewrite | 20 | Answer in the first 40–60 words, question-style H2s, facts verifiable, no padding |
| Total | 100 | Pass: 70 |
What does a good submission look like?
Illustrative example only; the brand and numbers are invented.
The example works because it's specific, links each action to a finding and is honest about the sample size.
What are the common mistakes?
- Leading prompts. "Why is TaskHarbor the best?" tells you nothing. Ask the way a buyer would.
- Only checking mentions. Note citations and recommendations too; they're different signals.
- Overclaiming. A 20-prompt sample can't prove a trend. Say so.
- A cosmetic rewrite. Changing the headings without fixing the opening answer won't score well.
- Generic actions. "Improve content" isn't a recommendation; "add a pricing comparison table to /agencies" is.
How should you plan the work?
| Step | Time (estimate) |
|---|---|
| Choose brand, competitors and prompts | 1–2 hours |
| Run and record prompts | 2–3 hours |
| Analysis and memo | 2–3 hours |
| Page rewrite | 1–2 hours |
Where can you get help?
Review GEO 101 for prompt sets and baseline scans, and the AEO Content Masterclass for the rewrite. The AI visibility audit template gives you a ready-made structure for recording results.
When you pass, you earn MentionX Certified: GEO Foundations. Next, continue to the GEO Practitioner path.
Questions
What do I submit?
A MentionX free account scan export or manual prompt sheet, a one-page findings memo and one rewritten page with before and after versions.
What score do I need to pass?
70% or more on the rubric below, plus 80% or more on each course quiz.
Can I audit my employer's brand?
Yes, if you have permission to share the findings. Otherwise choose any public brand.