Case: Known on Google, Invisible to AI

Anonymized at the client's request. Vertical paraphrased, numbers real.

The gap: recognized vs cited

We ran 12 local-intent questions ("best [service] near [area]?", "where to try [activity] for beginners?"). AI mentioned the brand in 4 of 12 answer sets — it knew the name from a directory — but cited its own site in only 2 of 36 engine runs. Recognition ≠ citation. AI knew the entity, not the offer.

Root cause: the offer hid behind JS

Fetching the homepage as GPTBot returned ~8KB vs ~210KB as a real browser. The entire services/pricing/menu block was client-rendered. AI saw the address and a hero sentence — not what to buy.

curl -A GPTBot https://example.com/ | wc -c  # 8,342
curl -A "Mozilla/5.0 ..." https://example.com/ | wc -c  # 210,411

robots.txt was clean; no noindex. The problem was purely content shape.

The fix list

1. SSR the offer for top 5 URLs (1 day): move service names, price, availability, and a 50-word “what we do” into initial HTML. This is the same 15+13 checklist from the GEO Score tool.
2. Add a quotable FAQ block (30 min): “What’s the best [service] for beginners near [area]?” — 45 words AI can lift.
3. Get onto the directory AI already cites (1 week outreach): the Perplexity-cited directory is open for listings; submission takes 10 minutes.

Takeaway

High Google rank and high AI citation are different games. One needs authority; the other needs quotability — a concise, structured answer inside the initial HTML. Our GEO Score checks the second half in 30 seconds; the crawlability check covers the first.