Get Started
Start here: buyers now ask AI first · Lesson 02

Decompose the answer-engine system

Break AEO into the parts you can test: crawl, render, understand, trust, cite, recommend, click and convert.

27 minWorked exampleField assignment
0%course
What you will leave with

A one-page diagnosis that names the exact broken stage and the next test.

“AI does not recommend us” is not a diagnosis. It is the end result of a chain. Good AEO work breaks that chain apart until you find the part you can prove, fix and measure.

The AEO chain

CrawlRenderUnderstandTrustCiteRecommendClickConvert

If any link breaks, the outcome looks the same: you are absent. But the fix is completely different. A crawler block is not solved with better copy. Weak evidence is not solved with llms.txt. A reputation problem is not solved with schema.

Worked diagnosis: SaaS page absent from ChatGPT

LinkQuestionEvidenceLikely fix
CrawlDid the bot fetch it?OAI-SearchBot: 200Not the blocker
RenderIs answer text in HTML?Pricing table loads with JS onlyServer-render key facts
TrustWho confirms the claim?No third-party sourcesAdd source ecosystem work
CiteDoes AI link you or others?Mentions brand, cites G2Improve first-party evidence and G2 profile

Fact or assumption?

Audit each belief before fixing

“AI can crawl us”Fact only if logs show verified bot requests returning usable responses.
“Our page answers it”Fact only if the answer is visible in server HTML under a matching heading.
“We are trusted”Fact only if answer engines or third-party sources already associate you with the claim.
“The change worked”Fact only if the same prompt and engine set changed after the dated fix.

The DARE prompt for AEO work

Act as a skeptical AEO analyst.
Decompose this visibility problem into: crawl, render, understand, trust, cite, recommend, click and convert.
For each stage, label what we know as fact, what we are assuming, and the cheapest test that would prove or disprove it.
Do not recommend a fix until the broken stage is identified.
Site/page:
Tracked prompt:
Current AI answer:
Available evidence:

Pick the right next move

Broken at crawlFix robots, WAF, rate limits, 403/429/5xx errors and verified crawler access.
Broken at renderMove answer-critical text into server HTML; do not hide it behind scripts or tabs only.
Broken at trustAdd original evidence, methodology, author credibility, reviews and source coverage.
Broken at conversionAI may be citing you already; measure referral sessions, leads and revenue before rewriting.
Do not let AI skip diagnosis.

AI is eager to solve. Force it to decompose first, audit assumptions second, and only then propose tests.

CCiteTrack AIv1.79.66
Where this shows up in the plugin
HomeTasksVisibilitySourcesCrawlersOptimizeAnalyticsAttributionReports

Turn the broken stage into a task

CiteTrack connects the answer, page evidence, cited sources and owner so the team can fix one specific stage instead of debating a vague visibility problem.

Open tasks18visibility work
High priority3needs owner
Evidence saved100%task detail
Resolved31verified later

/features/ai-analytics/Render issueMove key facts into HTML

/pricing/Claim driftUpdate pricing FAQ

/comparison/citetrack-vs-saas/Source gapImprove cited evidence

/demo/Conversion gapReview attribution

Use it in this lesson

Open the task created from the failed check.

Read the saved prompt, answer, page evidence and source rows.

Assign the fix with an expected outcome and review date.

Do the work

Field assignment

Scenario

Your team says “AI does not recommend us.” That statement is too big to fix. Break it into the smallest testable parts.

  1. Write the full chain from crawler access to conversion.

  2. Mark which links in the chain are facts and which are assumptions.

  3. Choose the cheapest test for the weakest link.

Definition of done

A one-page diagnosis that names the exact broken stage and the next test.