In late March, the founder of Christmas or Not ran our free scan on a whim. The result was a 24 — bottom-decile, labeled "Critical." ChatGPT could not state the company's pricing. Gemini placed the founding year four years too early. Perplexity confused the product with an unrelated holiday e-commerce site.
Thirty days later the score read 78. No agency, no paid placement, no product rewrite. Just a sequenced AEO playbook, executed in-house. This is the short version; the full case study has the receipts.
The diagnosis
The free scan surfaced six failure modes. The thing that mattered most: none of them were "the product is bad." Every failure was a parsing failure. The models didn't dislike Christmas or Not — they couldn't read it.
- Pricing lived in "Contact us" boxes with no schema.
- The founding date came from an old, wrong press mention.
- The customer list was trapped in PDF case studies.
- The product category parsed two different ways depending on the model.
The leverage move
When you have six AEO failures, you don't fix six things. You fix the two that compound. Schema fixed pricing, products, and founding in one shot. A forty-line llms.txt fixed parsing for every query at once. Both shipped in week one.
"We thought our problem was that AI didn't know us. Our actual problem was that AI couldn't parse our website. Once we fixed the parsing, three models started telling consistent stories about us in under a month."
Why they kept paying
The score hit 78 by day 32. Then, in mid-April, a new model shipped and briefly knocked them back down until the index caught up. Their weekly scan caught it within 48 hours. One schema tweak and the score returned.
That's the operational reality of AEO: it's not a one-time fix. The models keep moving. Your facts stay still. The gap between "true" and "what AI says" reopens unless something is watching. The full playbook — week by week, fix by fix — is in the Christmas or Not case study.