One person, one question:
does the answer cite you?

Search is splitting in two. One half still looks like Google's ten blue links, and the whole SEO industry knows how to measure that. The other half is a synthesized answer with a short list of sources underneath, and almost nobody can tell you whether they are on that list.

That second half is the gap b/cited was built to close. Ranking and being cited are different outcomes with different inputs, and a page can do well at one while being invisible in the other. If you cannot measure the difference, you cannot act on it.

Two signals, side by side, because neither is sufficient alone:

• Citation rate in AI answer engines. Tracked prompts are run against ChatGPT, Claude, and Perplexity, and the answers are parsed for whether your domain is named, whether it is cited as a source, and where in the list it lands. Those become mention rate, citation rate, and brand rank.
• Topical authority from Google Search Console. Ninety days of query data, read through the read-only webmasters.readonly scope, is clustered into topics and scored for how completely you own each one.

The gap between the two is the useful part. Owning a topic in Search Console while never being cited on it is a specific, fixable problem — and it is the one the content briefs are generated against.

The readiness scanner is the part you can check without an account, so it is the part worth explaining plainly. It fetches a URL and evaluates 30+ signals across nine categories: indexing, metadata, social cards, structured data, AEO signals (llms.txt, ai.txt, AI crawler stance, FAQ and author schema), international, mobile, security headers, and Core Web Vitals via Google PageSpeed Insights.

Every check is a rule about something observable in the response. There is no model judging your site and no hidden weighting — a failed check corresponds to a specific missing or malformed thing, which is why the output is a fix list rather than a verdict.

The public leaderboard runs that same scanner weekly against a curated set of domains per category. It is public on purpose: a scoring system nobody can inspect is not worth trusting, and publishing the results against well-known sites is the most direct way to show what the score does and does not capture.

AI citation is not a ranking you can reverse-engineer to a formula. Engines re-ingest training data and live indexes on their own cadences, so the realistic lag between a content change and a measurable shift in citation rate is weeks, not days. Anything promising faster is selling certainty that does not exist.

Tracked prompts are also a sample, not a census. They tell you how you do on the questions you chose to watch — which is actionable, and which is not the same as total visibility across everything anyone might ask.

Cloudflare-native end to end: Workers, D1, Vectorize, Queues, and Workflows, with an AI Gateway in front of the model providers. Running entirely at the edge is what keeps a solo-operated product cheap enough to have a free tier and simple enough to stay reliable without a team on call.

Search Console access is read-only and used for analysis and brief generation only. Your data is never used to train models. The details are in the privacy policy and the disclosures.