AEO Case Study: What Changed in 90 Days (Before and After)
Most AEO advice stops at the checklist. This is what the work actually looks like over 90 days — the diagnosis, the fixes in the order they were made, and what changed in how AI answer engines described the business. The numbers below come from a single mid-market services company we tracked prompt-by-prompt; the pattern repeats often enough to be useful as a template.
The starting point: invisible, not penalized
The company ranked on page one of Google for its main service-plus-city term and got steady organic leads. But when we asked the major engines the questions its buyers actually ask — “who are the best providers of X near me,” “compare the top X companies,” “is X worth it” — the brand was named in 2 of 30 prompts. Three competitors, all with weaker classic SEO, were named in more than twenty.
That gap is the whole point of answer engine optimization. Nothing was broken. The site simply wasn't quotable, and it wasn't present in the third-party evidence the models lean on.
Diagnosis: four concrete problems
| What we found | Why it suppressed AI mentions |
|---|---|
| Service pages written as brochure copy — adjectives, no facts | No sentence could be lifted as an answer, so nothing got cited |
| No pricing, no eligibility, no timelines anywhere on site | Buyer questions were answered by competitors who did publish specifics |
| Thin schema: only a bare Organization block, no service or FAQ markup | Machines had to guess what the business does and who it serves |
| Inconsistent name/address/category across directories and review sites | Entity ambiguity — the model couldn't confidently resolve one business |
The work, in the order it was done
Sequence matters. Fixing the crawlable, factual layer first means every later citation points at something worth quoting.
- Weeks 1–2 — extractable rewrites. Every service page got a 40-word direct answer under the H1, then question-shaped H2s. Marketing language moved below the facts.
- Weeks 2–4 — publish the specifics. Real price ranges, who the service is and isn't for, turnaround times, service-area list. This single change did more than anything else.
- Weeks 3–5 — structured data. Service, FAQPage and Organization markup with consistent identifiers, per schema markup for AI.
- Weeks 4–8 — entity cleanup. Same legal name, phone, and primary category across the map profile, the industry directories, and the two review sites the models kept quoting in this category.
- Weeks 6–12 — evidence, not backlinks. Two comparison pages ("us vs the common alternative"), a methodology page, and outreach to get the brand into three existing “best of” roundups that already ranked.
What changed by day 90
| Measure | Day 0 | Day 90 |
|---|---|---|
| Buyer prompts where the brand is named | 2 / 30 | 19 / 30 |
| Prompts where it appears in the top three named options | 0 | 11 |
| Answers citing the brand's own pages as a source | 1 | 14 |
| Descriptions containing a factual error about the business | 6 | 1 |
Note the last row. Half the value wasn't new visibility — it was correcting what engines already said. Wrong service areas and outdated claims were quietly costing qualified conversations.
Four transferable lessons
- Publishing specifics beats publishing volume. One page with real prices and eligibility outperformed a quarter of blog posts.
- Classic rankings don't carry over. Page-one SEO is not a defense against being left out of an answer. See AEO vs SEO.
- Consistency is a ranking factor for machines. Ambiguous entities get skipped in favor of unambiguous ones.
- Movement takes 4–8 weeks to show. Nothing changed in the first month. Judging AEO work at day 30 is how good programs get killed early.
Run the same diagnosis on your own brand
The playbook is repeatable: write down the 20–30 questions your buyers ask, ask the engines, record who gets named, then fix the extractable and entity layers before chasing anything else. The full sequence lives in how to do AEO and the condensed version in the AEO checklist.
If you'd rather see your own day-zero baseline first, Rooster's free AI Visibility Scan asks the engines your buyer questions and reports exactly where you're named, where you're missing, and what's being said wrong — the same starting picture this case study began with. The AI Visibility Tracker then watches those prompts week over week, so the day-90 column writes itself.