The short answer
A practical 90-day AEO program starts with a frozen buyer-question baseline, fixes technical and factual blockers, improves a few decision-critical pages, earns independent evidence, and measures the same prompts and business outcomes again.
Business outcome
You finish with fewer critical errors, stronger buyer evidence, a repeatable operating system, and a defensible view of what changed.
The process
Build it in five passes
Days 1 to 15: baseline and triage
Choose one category decision, map 12 to 20 questions, capture one consistent observation per prompt and engine, review referrals and platform citation data, inventory crawler access, and identify material answer errors. Rank problems by buyer impact and evidence, not by novelty.
Days 16 to 30: access and truth
Fix unintended crawler, rendering, canonical, indexing, sitemap, and internal-link issues. Build a source-of-truth matrix for high-risk product or business facts and correct owned contradictions. Keep training policy separate from search access.
Days 31 to 60: decision pages and proof
Improve two to four pages that map to the highest-value prompt and citation gaps. Lead with direct answers, explicit criteria, current facts, first-hand evidence, limitations, and useful artifacts. Align relevant structured data with visible content, without treating it as special AI markup.
Days 61 to 75: earn and distribute
Publish one defensible research, tool, dataset, benchmark, or expert contribution that deserves independent reference. Notify participating engines of material changes, maintain the sitemap, and conduct small, relevant outreach based on editorial fit.
Days 76 to 90: matched retest and decision
Repeat the baseline with the same core questions and settings. Compare recommendation share, citations, accuracy, referrals, business outcomes, index coverage, and cited pages. Document volatility and decide which interventions to maintain, expand, revise, or stop.
Before it ships
Quality checklist
- The program is scoped to one category decision and named buyer.
- Baseline prompts, engines, settings, and scoring are frozen before changes.
- Technical access and high-risk factual conflicts are fixed before content scale.
- Each new page solves a measured decision or evidence gap.
- At least one independent-worthy evidence asset has a real distribution plan.
- The day-90 review uses matched observations and separate business outcomes.
Copyable artifact
90-day operating board
Keep the plan deliberately small. Every task must connect to a buyer question, factual risk, access blocker, evidence gap, or measurement need.
PROGRAM: [category + buyer + market] OWNER: [name] START / END: [dates] BASELINE: [link] DAYS 1-15 - MEASURE - [ ] Buyer prompt map frozen - [ ] Baseline observations complete - [ ] Citation and referral views captured - [ ] Top factual risks named DAYS 16-30 - ACCESS + TRUTH - [ ] Crawler policy reviewed - [ ] Rendering, canonical, sitemap, and link blockers fixed - [ ] Source of truth matrix complete for high-risk facts DAYS 31-60 - PAGES + PROOF - [ ] Priority page 1 tied to prompt/gap: [URL] - [ ] Priority page 2 tied to prompt/gap: [URL] - [ ] Structured data aligned where relevant DAYS 61-75 - EARN + DISTRIBUTE - [ ] Evidence asset published: [URL] - [ ] Relevant outreach list and contribution prepared - [ ] Change notifications and sitemap verified DAYS 76-90 - RETEST - [ ] Matched baseline repeated - [ ] Recommendation, citation, accuracy, referral, and business metrics reviewed - [ ] Continue / revise / stop decisions recorded
Validation
How you know it is ready
- 01Every planned output links to a measured gap or business decision.
- 02No task assumes schema, `llms.txt`, or crawler permission guarantees recommendation.
- 03The final review can separate completed work, observed movement, and unknown causality.
Do not overclaim
Ninety days can establish evidence, fix controllable problems, and reveal direction. It cannot guarantee model refreshes, citations, rankings, or revenue, and should not be used to claim causation from a volatile before-and-after snapshot.
Questions
What teams usually ask
What should happen first in AEO?
Freeze the buyer-question baseline and fix high-risk access or truth problems. Publishing at scale before measurement makes the result hard to interpret.
How many pages should be created in 90 days?
Usually only a few decision-critical pages plus one strong evidence asset. Improve the smallest set that solves measured gaps.
When should the program expand?
Expand after the method is repeatable, the initial pages are maintained, and the review shows a clear next category, audience, or evidence gap worth funding.
Sources reviewed