Build a repeatable AI-visibility research program and interpret sampled observations honestly. Work through the four lessons, apply each exercise to a real, approved, personal, or sandbox site, and retain the result in your continuing portfolio.
Recommended prerequisite: access to a real, client-approved, personal, or sandbox website. Units 1–3 require no prior SEO experience.
Progress is self-reported. Mark a lesson complete only when you are satisfied with the applied work—not because you reached the bottom of the page.
Want completion marks to survive navigation? Save progress in this browser. Without opt-in, progress lasts only on the current page.
Lesson 11.1Not started
Building a Prompt Set
Prompt families, modifiers, buyer stages, fan-out questions, persona context, and prompt-demand uncertainty.
Open on YouTube. Interfaces, platform policies, and product details may have changed; learn the durable principle rather than memorizing buttons.
Learn
The working idea
Prompt families, modifiers, buyer stages, fan-out questions, persona context, and prompt-demand uncertainty.
Translate this principle into an observable decision: identify the audience task, the evidence available, the business consequence, and the verification method. Keep facts, assumptions, and recommendations separate.
Do the work
Exercise
Build a 30-prompt set across education, comparison, recommendation, troubleshooting, alternatives, and purchase. Record why each matters and whether real demand is known, inferred, or unknown.
Acceptance check: your output should name its source evidence, assumptions, intended user or business outcome, and how the conclusion could be verified.
Quick check
Compare your reasoning
1. Why should synthetic prompts not be labeled as measured demand?
Compare with guidance
They may not reflect actual user frequency or wording.
2. What makes a prompt set strategically representative?
Compare with guidance
Coverage of real customer tasks, stages, products, competitors, geographies, and known language.
Portfolio handoff
File the evidence
Save the exercise output in your own course workbook or project folder. This site does not upload or retain your work.
Lesson 11.2Not started
A Controlled Prompt-Test Protocol
Engine, model, account state, geography, date, personalization, repetition, and result capture.
Assignment note: Revisit the prompt-research lesson and the cross-engine citation study.
Open on YouTube. Interfaces, platform policies, and product details may have changed; learn the durable principle rather than memorizing buttons.
Learn
The working idea
Engine, model, account state, geography, date, personalization, repetition, and result capture.
Translate this principle into an observable decision: identify the audience task, the evidence available, the business consequence, and the verification method. Keep facts, assumptions, and recommendations separate.
Do the work
Exercise
Run each of ten prompts at least three times on two approved surfaces. Record exact prompt, timestamp, engine/model, conditions, response, mentions, recommendations, citations, URLs, and variance.
Acceptance check: your output should name its source evidence, assumptions, intended user or business outcome, and how the conclusion could be verified.
Quick check
Compare your reasoning
1. Why are one-off screenshots weak evidence?
Compare with guidance
Outputs vary and cannot establish repeatability or prevalence.
2. Which conditions should be logged?
Compare with guidance
Prompt, engine/model, date/time, location, account/personalization state, run number, and observed sources.
Portfolio handoff
File the evidence
Save the exercise output in your own course workbook or project folder. This site does not upload or retain your work.
Lesson 11.3Not started
Mentions, Citations, Recommendations, and Sentiment
Mention versus linked citation versus favorable recommendation; source overlap; engine-specific indexes and retrieval.
Open on YouTube. Interfaces, platform policies, and product details may have changed; learn the durable principle rather than memorizing buttons.
Learn
The working idea
Mention versus linked citation versus favorable recommendation; source overlap; engine-specific indexes and retrieval.
Translate this principle into an observable decision: identify the audience task, the evidence available, the business consequence, and the verification method. Keep facts, assumptions, and recommendations separate.
Do the work
Exercise
Code 50 sampled outputs using a strict taxonomy: absent, mentioned, described, recommended, cited, linked, sentiment, position, and source type.
Acceptance check: your output should name its source evidence, assumptions, intended user or business outcome, and how the conclusion could be verified.
Quick check
Compare your reasoning
1. Can a brand be mentioned without being cited?
Compare with guidance
Yes.
2. Why should cross-engine results not be combined blindly?
Compare with guidance
Engines use different indexes, retrieval, models, contexts, and citation behavior.
Portfolio handoff
File the evidence
Save the exercise output in your own course workbook or project folder. This site does not upload or retain your work.
Lesson 11.4Not started
Measuring AI Visibility Without Pretending It Is Complete
Sampled share of voice, citations, referral traffic, logs, branded search, conversions, self-reported attribution, and product-report limitations.
Open on YouTube. Interfaces, platform policies, and product details may have changed; learn the durable principle rather than memorizing buttons.
Learn
The working idea
Sampled share of voice, citations, referral traffic, logs, branded search, conversions, self-reported attribution, and product-report limitations.
Translate this principle into an observable decision: identify the audience task, the evidence available, the business consequence, and the verification method. Keep facts, assumptions, and recommendations separate.
Do the work
Exercise
Design a dashboard with separate sections for eligibility, sampled visibility, source/citation presence, traffic, conversion, and qualitative customer evidence. Include caveats next to every metric.
Acceptance check: your output should name its source evidence, assumptions, intended user or business outcome, and how the conclusion could be verified.
Quick check
Compare your reasoning
1. Is vendor share of voice market-wide ground truth?
Compare with guidance
No; it is based on the vendor’s sampled prompts and definitions.
2. Why separate citation from conversion?
Compare with guidance
Being used as a source is not the same as sending a visit or influencing a purchase.
Portfolio handoff
File the evidence
Save the exercise output in your own course workbook or project folder. This site does not upload or retain your work.
Unit 11 exam
Apply it without pretending the browser is your professor.
Status: Not attempted
Knowledge prompts
Distinguish prompt from keyword; define mention, recommendation, and citation; list five test controls; explain sampled share of voice.
Practical assignment
Submit a 30-prompt library, repeated two-engine test, coded output dataset, and a one-page interpretation with limitations.