United States Research
CiteAngle tracks how brands are discovered, named, cited and converted across the US market, and every source and denominator stays explicit.
The problem
Nobody is checking whether the assistants name you.
Rank trackers watch blue links. When a buyer asks Gemini or ChatGPT instead, something outside your dashboard picks the brand that gets quoted.
How we fix it
We pick one industry and ask the same questions across every answer screen.
Your competitors get counted in the same run, from the same answers. That turns a score into a picture of the whole category.
What you get
You can see which ground is empty and which to take first.
The studies below came out of those runs. Read one and you will know the order of work for your own category.
See the state model, denominator rules, owned-platform evidence, controlled observations and the boundary between automated diagnosis, the publishing activation gate and expert-led external work.
Series · Side by Side: public-page comparisons
AI-visibility measurement products, lined up on the same axes using only what each vendor states on its own public pages. Every fact carries its access date, and a fact that changes later gets a correction block on the page.
Series · Verify Before You Buy: reading AI-visibility claims
One reporting pattern per part, turned into questions a buyer can ask in writing. It names practices, never vendors. Every quoted sales line carries a format-example label, and our own figures sit in the public claims registry.
Why one run cannot rank vendors, and seven runs can
Side by Side Public-page comparison
Eight global vendors, lined up on the axes their own public pages disclose: pricing, engines, repeats and confidence intervals, raw data, verification artifacts. Plus five questions to ask before you buy.
Market
Google and Bing now publish official AI exposure telemetry. Here is a market map of the US AI answer surfaces and what each one actually reports.
Demand
AI search breaks one query into parallel sub-questions. If you still measure demand in keywords, you are counting the wrong unit.
Sources & citations
Controlled benchmarks stress-tested dozens of popular citation tactics. What survived, what didn't, and where AI citation is really decided.
Sources & citations
One full pass over 50 US questions across all 15 answer surfaces: 8,662 citations, 1,987 domains, a head of ten that carries 27.2% of the total, and a tail of 1,088 domains cited exactly once.
Sources & citations
Academic preprints, a vendor's own blog and search-industry guidance now point the same way: one pass over an AI engine yields a point estimate, not a number you can budget on. Here is the published evidence, with links and dates.
Due Diligence
One deck says your AI mentions will triple. Before that number reaches a budget, four denominator questions decide whether it is a procurement-grade metric or a vanity figure, and published panel data shows the denominator itself moves. Verify Before You Buy, Part 1.
Action
Google and Bing already run free, official telemetry on your site. Stand these rails up first, then buy measurement that answers what they can't.
Market
OpenAI, Google, and others now sell ad space around AI answers. None of them sell the answer itself, and that changes what to measure before you spend.
Market
One listed company's filings show advertising falling from 29.16% of revenue to 17.46% over six years while the advertising line itself multiplied 4.16 times. The share moved because revenue moved faster.
Market
One listed company's advertising ran in phases: a push peaking at 35.08% of revenue in the only loss-making quarter of 29, then 13 quarters in a row where revenue outgrew advertising, then a second push.
Separate unaided category demand, reputation, comparison and purchase-stage questions.
Trace direct citations, retrieved sources, organic candidates, news and video occurrence.
Turn conflicting names, products, locations and claims into a verified action queue.
Register target, held-out and delayed cohorts before collection and bind change receipts to remeasurement.
Keep the national baseline and selected state denominators, location receipts and action queues distinct.
Study how repeatable AI-answer visibility measurement is: repeated runs, sampling variance and confidence-interval reporting, grounded in the published literature and held to our published protocol.
Market maps, baselines and verification reports live in the reports library. Each keeps a permanent URL, and we date every revision on the page. The measured figures themselves are downloadable as JSON from the open aggregates dataset.
Customer-owned Google and Bing aggregates, controlled AI-answer observations, and owned conversion records stay separate. Every finding points to one decision you can act on.
Read the research, then run it on your own brand
Every brief above ends in a measurable move. A free AI visibility check runs the same protocol on your own buyer questions, before you spend anything.
Ready to turn US visibility evidence into a market plan? Tell us the US buyer, target state and decision date.
Request a written proposal