We Don't Guess If AI Said Your Name. We Test the Reading.

Most tools that claim to track "AI visibility" are, underneath, guessing. They skim an answer, notice your name went by, and call it a mention. Ask them how sure they are and the honest answer is: not measured. We built AlphaCitation to do the opposite: put real buyer questions to the engines ourselves, read every answer, and keep receipts.

Most tools assume they heard your name. We test it.

Every scan checks whether your brand was actually named across real buyer questions: not a keyword scrape, a parser reading the answer the way a person would. We don't take that reading on trust. In July 2026 we tested it against a reference set of real engine answers that people had labelled by hand, in two rounds of review, so the right answer for each one was fixed before our system saw it.

The result on that set: zero errors. We publish the bound rather than the streak: at 95% confidence, that result puts detection accuracy at 97.3% or better. It is a result on a tested set, not a promise about every future answer, which is why every score we report carries an interval that allows for some error instead of assuming none.

A separate check used a different kind of grader: an independent AI judge, a separate model with no view of what our system had concluded. On mention detection it agreed with our reading on all but one answer.

Hearing your name is the easy part. Knowing the tea is the real work.

Being named is only half the story. The harder, more valuable question is: what did AI actually say about you when it said your name? Was it a warm recommendation, a neutral factual mention buried in a list, or a backhanded "sure, but..."? That's a genuinely harder problem than detecting a name. Sentiment is a judgment call, not a lookup, and we hold ourselves to the same bar: measured against an independent AI judge rather than a human panel, and never self-graded.

In our most recent measurement, in July 2026, our sentiment reading agreed with that judge 85.4% of the time, up from 43% when we first measured it earlier that month. The figure moves a few points between runs on the same set, so we publish the measurement rather than a rounded claim, because a stat you can't defend isn't worth having. We hold sentiment to the same standard as mention detection, and we'll keep publishing it as it improves.

Why this matters more than a score

Anyone can show you a dashboard. Very few will show you how they know the dashboard is right. That's the difference between a tool that tells you AI mentioned you, and one that can show how it knows, and then tell you what was actually said.

Full technical detail lives on the methodology page. If you want to know what AI is actually saying about your brand, run a free scan.