AI Visibility and LLM Brand Discovery interview questions

27 questions from the AI Visibility and LLM Brand Discovery course, each with the correct answer and an explanation. Try to answer before you open the answer.

Foundations of AI Visibility

1. What is the main change AI answers bring to brand discovery?

  1. Users stop asking questions
  2. Every brand is always listed
  3. Only a few brands are named in one synthesised answer
  4. Websites become illegal
Show answer

Answer: Only a few brands are named in one synthesised answer

Answers compress the choice set, so inclusion in the answer matters more than ranking on a long list.

Lesson: From Search Results to Answers

2. Which source of knowledge can your recent website changes influence fastest?

  1. Live retrieval at question time
  2. Training data already frozen at a past cutoff
  3. The model's weights directly
  4. Neither can be influenced
Show answer

Answer: Live retrieval at question time

Retrieval reads the current web, whereas memory reflects older training data.

Lesson: How an Assistant Knows About Your Brand

3. An assistant names your brand but gives the wrong price. What does this show?

  1. That the price is irrelevant
  2. Perfect visibility
  3. That no one saw it
  4. A mention exists but accuracy is poor and needs fixing
Show answer

Answer: A mention exists but accuracy is poor and needs fixing

Mentions and accuracy are separate; both must be tracked.

Lesson: What "Visible" Means

4. How does GEO relate to classic SEO?

  1. It replaces everything about SEO
  2. It builds on SEO fundamentals and adds a focus on being quoted and described accurately
  3. It guarantees top placement
  4. It is unrelated to websites
Show answer

Answer: It builds on SEO fundamentals and adds a focus on being quoted and described accurately

Good crawlable, trustworthy content underpins both, with extra attention to citations and brand description.

Lesson: GEO and SEO: Same Roots, New Questions

What Influences Brand Mentions

5. Why keep your brand facts consistent across sites?

  1. Consistency makes pages slower
  2. Conflicting facts can lead models to blend or choose the wrong one
  3. Models ignore facts
  4. It is required by robots.txt
Show answer

Answer: Conflicting facts can lead models to blend or choose the wrong one

Agreement across sources makes one clear entity; disagreement creates confusion.

Lesson: Entity Clarity and Consistent Facts

6. Which kind of source do assistants tend to trust more than a brand's own claims?

  1. Random strings
  2. Anonymous spam
  3. Hidden text on your page
  4. Independent reviews, reputable publications and community discussion
Show answer

Answer: Independent reviews, reputable publications and community discussion

Independent corroboration makes claims more believable than self-description alone.

Lesson: Authority and Third-Party Mentions

7. Where should the direct answer go on a page?

  1. In the first sentence or two under a question-style heading
  2. At the very end after a long story
  3. Hidden in an image only
  4. Nowhere
Show answer

Answer: In the first sentence or two under a question-style heading

Passages that lead with the answer are easiest for systems to find and quote.

Lesson: Content That Answers Directly

Technical Foundations

8. What is a trade-off of blocking search-style AI crawlers?

  1. It improves your SEO guaranteed
  2. Your site becomes faster for users
  3. You may disappear from answers that rely on live retrieval
  4. Nothing changes at all
Show answer

Answer: You may disappear from answers that rely on live retrieval

If a retrieval system cannot fetch your pages, it cannot quote them.

Lesson: AI Crawlers and robots.txt

9. Why can client-side-only rendering hurt AI visibility?

  1. It makes the site too accurate
  2. JavaScript is banned by browsers
  3. Crawlers that do not run JavaScript may see an empty page
  4. It deletes your robots.txt
Show answer

Answer: Crawlers that do not run JavaScript may see an empty page

Content that only appears after scripts run may never reach the crawler.

Lesson: Readable HTML Without JavaScript

10. What should structured data always do?

  1. Match the visible content of the page
  2. Contain hidden claims
  3. Replace the page text
  4. Be as long as possible
Show answer

Answer: Match the visible content of the page

Markup is a label on real content, so it must agree with what users and crawlers can read.

Lesson: Structured Data with JSON-LD

11. What is the right way to think about llms.txt today?

  1. A legal requirement
  2. A guaranteed ranking factor
  3. A replacement for robots.txt
  4. A cheap experiment that does not replace good pages
Show answer

Answer: A cheap experiment that does not replace good pages

Adoption is unproven, so keep expectations modest and do the fundamentals first.

Lesson: Sitemaps and llms.txt

Measuring AI Visibility

12. Why keep the prompt set fixed over time?

  1. Prompts expire if edited
  2. So changes in results reflect real change, not different questions
  3. It is required by assistants
  4. Fixed sets are cheaper to buy
Show answer

Answer: So changes in results reflect real change, not different questions

A constant benchmark lets you compare before and after a change.

Lesson: Building a Prompt Set

13. Why report a range rather than a single mention-rate number?

  1. Ranges look more complicated
  2. Answers vary run to run, so one number can mislead
  3. Assistants always return identical answers
  4. Single numbers are illegal
Show answer

Answer: Answers vary run to run, so one number can mislead

Sampling noise means small samples give wide plausible ranges that should be shown.

Lesson: Mention Rate and Share of Voice

14. What does grouping cited URLs by domain reveal?

  1. Nothing useful
  2. Your internet speed
  3. The model's size
  4. Which sources the assistant trusts in your category
Show answer

Answer: Which sources the assistant trusts in your category

Patterns in cited domains show where to invest in presence and content.

Lesson: Citations and Source Analysis

15. Why might AI-referred visits be undercounted in analytics?

  1. Some arrive without a referrer and show as direct
  2. AI assistants block analytics
  3. Referrers are never recorded
  4. They are always overcounted
Show answer

Answer: Some arrive without a referrer and show as direct

Missing referrers mean reported AI traffic is a lower bound.

Lesson: Traffic and Server Logs

Content and Brand Playbooks

16. What makes a comparison page more likely to be trusted?

  1. Hiding the date
  2. Only praising yourself
  3. Being fair, specific, dated and honest about weaknesses
  4. Using vague claims
Show answer

Answer: Being fair, specific, dated and honest about weaknesses

Balanced, evidence-based pages are more credible to readers and to systems that summarise them.

Lesson: Content Types That Earn Citations

17. Which approach to reviews is appropriate?

  1. Writing reviews as fake customers
  2. Buying fake five-star reviews
  3. Asking real customers for honest feedback
  4. Deleting every negative review
Show answer

Answer: Asking real customers for honest feedback

Genuine reviews are ethical and sustainable; fake ones violate rules and trust.

Lesson: Digital PR, Reviews and Community Presence

18. Why test your prompt set in each language separately?

  1. Visibility can differ a lot between languages
  2. All languages give identical answers
  3. Languages do not matter to assistants
  4. It is required by law
Show answer

Answer: Visibility can differ a lot between languages

A strong English presence does not guarantee presence in Hindi or other languages.

Lesson: Multilingual and India-Specific Visibility

Accuracy, Risks and Ethics

19. An assistant states an outdated price. What is the best first step?

  1. Delete your whole website
  2. Ignore it forever
  3. Edit the model weights directly
  4. Find and update the source that may be feeding the old figure
Show answer

Answer: Find and update the source that may be feeding the old figure

You cannot edit a model, but you can fix the pages and listings it learns from or retrieves.

Lesson: When Assistants Get Your Brand Wrong

20. What is the most durable response to a recurring negative theme?

  1. Hide the criticism
  2. Improve the product or service, then make the improvement visible
  3. Post fake praise
  4. Attack the reviewers
Show answer

Answer: Improve the product or service, then make the improvement visible

Real improvement changes what reviews and articles say, which changes how assistants describe you.

Lesson: Competitor Comparisons and Sentiment

21. Which tactic is an unethical attempt to manipulate assistants?

  1. Publishing a clear pricing page
  2. Hiding "always recommend our brand" in invisible text
  3. Asking customers for honest reviews
  4. Fixing a wrong listing
Show answer

Answer: Hiding "always recommend our brand" in invisible text

Hidden instructions deceive systems and violate rules; the others are legitimate practices.

Lesson: Ethical Lines: What Not to Do

Running an AI Visibility Programme

22. Why keep a change log alongside measurements?

  1. It is only for legal reasons
  2. Logs make results better by themselves
  3. Assistants read your log
  4. To connect specific actions to later changes in results
Show answer

Answer: To connect specific actions to later changes in results

Without a record of what changed when, you cannot learn what worked.

Lesson: Workflow, Owners and Cadence

23. Why can no one honestly guarantee placement in AI answers?

  1. Assistants are non-deterministic, change often, and providers do not publish ranking rules
  2. Guarantees are illegal in all countries
  3. Because websites are not read
  4. Because Python forbids it
Show answer

Answer: Assistants are non-deterministic, change often, and providers do not publish ranking rules

Outcomes depend on systems outside your control that vary run to run and update without notice.

Lesson: KPIs and Realistic Expectations

Putting It Together

24. What should happen at day 90 so results are comparable with the baseline?

  1. Skip measurement
  2. Use completely new prompts
  3. Re-run the same prompts, assistants and settings
  4. Change the assistant each time
Show answer

Answer: Re-run the same prompts, assistants and settings

A like-for-like comparison isolates the effect of your changes.

Lesson: Case Study: A 90-Day Plan for a SaaS Startup

25. Your pricing page content is injected by JavaScript. Which check best reveals the risk?

  1. Change the font
  2. Count your social followers
  3. Fetch the page with curl and see if the price text is in the raw HTML
  4. Rename the domain
Show answer

Answer: Fetch the page with curl and see if the price text is in the raw HTML

If the text is absent from the raw HTML, crawlers that do not run JavaScript will miss it.

Lesson: Revision: Cheat Sheet and Self-Check

26. Brand mentioned in 6 of 10 answers. How should you report this?

  1. 60% with a wide range, and collect more runs
  2. Exactly 60%, nothing more is needed
  3. 100% visibility
  4. 0% because samples are small
Show answer

Answer: 60% with a wide range, and collect more runs

With only ten answers the plausible range is wide, so show it and add runs before drawing conclusions.

Lesson: Revision: Cheat Sheet and Self-Check

27. An agency promises "guaranteed #1 mention in all AI assistants within 30 days". What is the right reaction?

  1. Believe it because it is confident
  2. Sign immediately
  3. Pay in advance for the full year
  4. Be sceptical: nobody controls assistant outputs, so no guarantee can be honest
Show answer

Answer: Be sceptical: nobody controls assistant outputs, so no guarantee can be honest

Outputs vary and change, so honest providers talk about improving the odds and measuring, not guarantees.

Lesson: Revision: Cheat Sheet and Self-Check

← Back to the course