AI visibility score
We sell a number that moves
Score = mention rate × 50 + citation rate × 30 + position × 20, out of 100. The question set is agreed and locked at kickoff and only changes in writing. Every raw answer is archived and available to you on request. Movement within ±5 points is treated as noise, not a result.
180 measurements a month
20 questions a real customer would ask × ChatGPT, Gemini, and Perplexity × three runs each. Three runs matter: generative models vary, so a mention counts as stable only if it appears at least twice.
Mention rate (weight 50)
The share of questions where you are mentioned reliably. A low score means the AI has nothing about you to talk about, so content is the fix.
Citation rate (weight 30)
How often a mention comes with a link to your page. Being named without a source is exposure, not traffic. Structured data and verifiable facts fix it.
Position (weight 20)
Where you appear in the answer. Present but not first means weak authority signals.
Process
The order is not negotiable
Technical audit
About five working days. Full crawl, plus fetches as OAI-SearchBot, Claude-SearchBot, and PerplexityBot, the crawlers that decide search citations rather than the training crawlers. You get findings in three tiers and an assessment of what your platform can and cannot do. The fee is credited in full against a follow-on project.
Technical optimisation
Fixed-price modules chosen from the audit findings, not billed by the hour. Delivery includes a free baseline AI visibility measurement.
Baseline
Measure the starting score once the technical work is right, so later movement means something.
Subscription, if you want it
Rules change, competitors move, content goes stale. The technical work stands on its own; you can wait three months before deciding.
Technical
What the technical work covers
Structured data framework
Head injection, template hooks, and the base for per-page markup.
JSON-LD entities
Organization, LocalBusiness, and Product markup, including certifications, contacts, and locations.
H1 in server-rendered HTML
Headings that exist in the source, not only after JavaScript runs, so crawlers can read them.
hreflang and locale mapping
Correct language and region signals, so translations are not read as duplicates.
Validation and AI crawler testing
Rich Results Test, schema validators, and real fetches by AI crawlers.
Split delivery
We write the code. For meta text, image naming, and back-office work we give you the rules, templates, and one review pass, and your team applies them at a fraction of the cost.
Subscription
Subscription tiers
Prices are in US dollars, converted at NT$32 to US$1; Taiwan clients are invoiced in New Taiwan dollars, and other currencies are quoted on request. Subscriptions exclude website build and hosting: your site stays where it is. Score targets are targets, not guarantees. If the score misses its growth path for three consecutive months, we raise it ourselves and attribute the cause honestly.
Stable (L1)
The promise is that the score does not slip. Quarterly measurement, monthly automated technical health checks, proactive alerts, quarterly fact checks. No new content, so the mention rate will not rise.
Advanced (L2)
The promise is that the score goes up. Monthly measurement, quarterly competitor comparison, one SEO / GEO article a month, quarterly keyword and content-gap reporting.
Leading (L3)
The promise is that you lead your competitors. Monthly measurement against three competitors, four articles a month with visuals, monthly fact checks, a monthly strategy session, and a named contact.
FAQ
Is a paid audit worth it? Free tools exist.
Free tools tell you there are 47 issues. They do not tell you which three actually block AI from reading the site and which 44 change nothing. You get a ranked list with fixed prices per fix, and the fee is credited against the project. The list is yours either way.
Do we have to take the subscription afterwards?
No. The technical work is a complete project. The results and the baseline report are yours.
You calculate the score yourselves. How do we know it is honest?
The question set is locked and only changes with your written agreement, every raw answer is archived for you to inspect, and competitors are measured with the same questions and the same method. You can also ask any question yourself and check.
Can you guarantee first place?
No, and nobody who does should be believed. We guarantee measurable things: validated structured data, content that AI crawlers can actually fetch, healthy indexing, and a monthly score with the raw data behind it.
Our platform is restrictive. Is this possible at all?
Start with the audit. Platform limits are identified there, before you spend anything on implementation. Most limits have a workaround; the ones that do not, we tell you about before quoting.
Can we subscribe without doing the technical work first?
You can, but we will not promise any score improvement. With the structure unfixed, most of the benefit of new content is absorbed before it reaches anyone, and we are not comfortable charging for that. In that case the subscription covers monitoring, recommendations and content only.