GEO KAMI scans your site for 37+ AI-readiness signals, then measures how ChatGPT, Gemini, Claude, Perplexity and Google AI actually mention, cite and recommend your brand.
No credit card required · Scanning starts after you sign in
The workspace
One board for the whole question: how the engines describe you, how your site scores on 37+ readiness checks, what Google Search Console and Analytics say is happening, and where you sit against your industry.
Mentioned
of 300 sampled answers
Engine coverage
Sources winning your citations
Questions shaping your visibility
Best GEO platform for SaaS teams?
Recommended
How do I get cited by ChatGPT?
Mentioned
AI visibility tools comparison
Absent
The engines your buyers actually ask
GEO KAMI samples ChatGPT, Gemini, Claude, Perplexity and Qwen — each engine separately, never blended into one score.
The architecture
Every source arrives on its own line. The funnel cross-references them, abstracts what matters and decides — then each decision leaves on its own line.
Each source is its own feed.
Your site
37+ readiness checks across seven categories
AI crawler telemetry
Which AI bots actually arrived, and what they were able to read
Google Search Console
Clicks, impressions and the queries behind them
Google Analytics 4
Sessions and conversions
Competitors
Side by side with the sites you name
Search demand
Rising interest in the topics around your category — relative, not absolute volume
Where the sources stop being separate.
Cross-source aggregation
Six feeds, aligned on the same site and the same window
Abstraction
Signals that matter, noise dropped
Analysis
What to fix first, and what to publish
GEO visibility UP, rank UP.
MCP & data sources
Produced when a bot cannot read something — the machine-readable surface that fixes it
Website GEO optimization & repair
Every failed check carries its fix; this is where they get applied
GEO/SEO-optimized content publishing
Pages written for both the crawler and the answer, not just the ranking
AI visibility monitoring
Which engines moved, and by how much, over time
GEO Forge plugin
the workerApplies the repairs and generates the well-known files on the site itself, and logs which AI bots actually arrive.
Method
AI answers are noisy, and a single number hides that. Every figure GEO KAMI shows carries the sample it came from.
Category questions × runs × engines. A count without its sample is not a measurement, so we never render one without the other.
When a classifier cannot tell whether you were recommended, the row is counted as borderline — never folded into the recommendation rate.
Per-model unit cost and month-to-date spend come from the usage the provider itself reports, and show as unknown rather than an estimate.
The shift
A search result was a list of options. An AI answer is a recommendation. When a model names three vendors, the fourth is not ranked lower — it is simply absent. GEO KAMI measures that answer directly, instead of inferring it from rankings.
One answer replaces ten blue links. If you are not in it, you had no placement at all.
Models recommend a handful of vendors. Selection happens inside the answer, before anyone visits a page.
We sample the real questions in your category and record what came back — not what a ranking model predicts.
Platform
37+ checks in seven categories: robots.txt and sitemap, llms.txt quality, markdown negotiation, AI bot rules, content signals, JSON-LD and security headers.
ChatGPT, Gemini, Claude, Perplexity and Google AI are sampled and reported separately. One blended score hides which engine ignores you.
MCP server cards, A2A agent cards, OAuth discovery, OpenAPI specs, WebMCP, and x402/ACP commerce signals — checked, not assumed.
Google Search Console and GA4 as first-class sources, plus a REST API for visibility summaries, run history and cost data.
Beyond the audit
Two questions the same collection already answers: are AI crawlers actually reaching you, and is your score normal for your category?
Crawler hits identified by name — GPTBot, ClaudeBot, PerplexityBot and the rest — instead of one anonymous “bot” line in your logs.
AI crawler visits · last 14 days
Identified from the GeoForge traffic log
Your score is placed against the p10, p50 and p90 of sites in the same category, so a number like 84 means something.
84
above 82% of SaaS sites
SaaS sites · sample 1,204
FAQ
What GEO KAMI measures, how it measures it, and what changed
GEO is the work of making your site readable and quotable by AI systems, so that when a model answers a question in your category it can find you, understand you, and name you. Traditional SEO competes for a position in a list; GEO competes for a place in the answer itself.
In practice yes — the two names describe the same work. AEO (Answer Engine Optimization) is the older term, from the era of featured snippets and voice assistants, when the goal was to be the single answer a search engine read out. GEO (Generative Engine Optimization) is the name that took over once AI systems began writing answers rather than extracting them. You will also see LLMO (Large Language Model Optimization), and simply "AI optimization". All of them describe one discipline: making a site readable and quotable enough that an AI system will name it, link to it, or recommend it. We say GEO because that is where the research literature and the tooling settled, but if you arrived searching for "AEO", you are in the right place.
SEO competes for position in a ranked list, so a page can rank well while the answer goes elsewhere. GEO competes for inclusion in a generated answer, where there is no second page. The signals overlap — crawlability, structure, authority — but the target and the measurement are not the same.
We fetch your site and run 37+ checks across seven categories: discoverability (robots.txt, sitemap, security.txt), accessibility (llms.txt, markdown negotiation, canonical), bot control, agent protocols, AI readability, security headers, and commerce signals. Every check returns a pass, a warning or a failure, with the evidence it was judged on.
Anonymous scanning was being abused — a free, unmetered endpoint that calls paid models is a standing invitation. Scans are now tied to your account, so your credits, results and history stay yours.
Visibility data describes how AI engines talk about a brand. Leaving it readable by anyone who types a domain would let a competitor pull another company's coverage. A domain that is not proven to be yours is not offered for measurement.
ChatGPT (OpenAI), Gemini (Google), Claude (Anthropic), Perplexity and Qwen, each sampled separately. Blending the engines into one score hides the most useful finding — the one engine that never mentions you.
Mentioned means your brand appeared in the answer. Cited means the model linked to your page as a source. Recommended means the answer actually put you forward as an option. They are counted separately because they are three different outcomes — being named without being linked is a different problem from being linked without being chosen.
Every count is shown with the sample behind it: questions × runs × engines. Cases the classifier cannot resolve are reported as borderline rather than folded into the recommendation rate. And the figures stay descriptive — we do not dress them up as a confidence score the data cannot support.
Yes. Connecting Google Search Console imports your properties and can confirm domain ownership automatically; GA4 summaries sit alongside your AI visibility data. Both are read-only.
Yes. A REST API exposes visibility summaries, run history and cost data under your account's API keys, so the same numbers can feed your own dashboards and reports.
Run the 37+ point audit, verify your domain, and see how five engines answer the questions that decide your category.
✓ No credit card required ✓ 37+ checks in one scan ✓ Verify ownership to unlock visibility
ChatGPT
Claude
Gemini
Perplexity
Grok
Qwen
