§ 02 — Service
Generative Engine Optimization
Getting a site into the answer, not just into the index. Three deliverables, all of which produce something you can check yourself.
What GEO actually is
GEO is the work of making a website reachable, readable and worth quoting by generative engines — ChatGPT, Claude, Perplexity, Gemini, and on the Chinese side Doubao, Kimi, DeepSeek and Yuanbao.
It overlaps with SEO but the target is different. SEO competes for a rank on a results page. GEO competes to be the source an answer is built from. A page can rank well and never be cited; a page can be cited without ranking at all.
The part most sites get wrong
The crawlers that decide whether you get quoted do not execute JavaScript. If your body text is rendered on the client, they see an empty shell. This is not a subtle optimisation problem — it is the difference between existing and not existing.
Two of our own sites, measured the same way, on the same day:
| Page | Raw HTML | Body text, JS disabled | Verdict |
|---|---|---|---|
sms.tempmail2026.xyz | 28,560 | 3,999 | Readable, thin |
mail.tempmail2026.xyz | 6,909 | 20 | A title and nothing else |
Method: fetch the page, strip <script> and <style>blocks entirely, strip remaining tags, unescape entities, collapse whitespace, count characters. Counting bytes instead of characters, or leaving script contents in, inflates the number several times over — we made exactly that mistake in an earlier draft and corrected it.
What you get
1 — Readiness audit · ¥1,980
A report, not a dashboard login. Per-crawler reachability tested with the real user agents. Body text measured with JavaScript disabled. Structured data coverage and validity. hreflang bidirectional consistency. Corpus density. Then a remediation list graded by impact.
2 — Technical remediation · from ¥5,800
Implementation of that list: robots.txt, static rendering or prerendering, JSON-LD, sitemap and hreflang, and restructuring content so a machine can lift a clean answer out of it.
3 — Citation monitoring · ¥399/month
A fixed query set run on a schedule against both English and Chinese engines, recording whether you get mentioned or cited, plus a monthly report of crawler hits from the server logs. Trend lines, not a single snapshot.
Three things worth correcting
Much of what circulates about GEO is either outdated or was never true. Three we check before charging anyone for it:
| Common advice | Current reality |
|---|---|
llms.txt is the key to GEO | In an Ahrefs sample of 137,000 domains, 28% had published one — and 97% of those files received zero requests. Worth adding because it costs nothing. Not worth paying for. |
| Pile on FAQPage schema | Google stopped showing FAQ rich results on 7 May 2026 and removed the documentation on 15 June. It still gives an LLM clean question–answer structure, so keep it — but stop counting it as a search win. |
| Add review stars to your own site | Self-serving reviews on Organization or LocalBusinessare not eligible. Product is the exception. We do not put stars on our own pages. |
What we do not promise
No promise that an engine will recommend you, place you in a top-N list, or cite you at any particular rate. Generative output is probabilistic and the model side is opaque. Nobody can guarantee a specific output, and a vendor who does is selling the old “guaranteed first page” trick with a new name on it.
What we do commit to is checkable: before-and-after crawler reachability, before-and-after body-text measurements, structured-data validation results, hreflang consistency. You can re-run every one of them yourself.
On timing, the honest version: technical changes can be verified within weeks. Training data and index refresh cycles are outside anyone's control, so citation rates have no fixed timetable.
Where to look next
- Case studies — including the 20-character page above, being rebuilt in public.
- Pricing — all three deliverables, plus website builds.
- Colophon — how this site is built, in full.