Built to be checked, not trusted.
Saylent is an independent product with one job: show you exactly how AI engines answer your buyers, and hand you the receipts, not a dashboard.
Why Saylent exists
Buyers stopped Googling. They ask ChatGPT, Claude, Gemini and Perplexity what to buy, and the engines answer with a short list of names. If yours isn't on it, nobody tells you. You just win fewer deals, with no idea why.
Most "AI visibility" tools sell you a subscription to watch a score move up and down. Saylent does the opposite: it gives you the actual answers your buyers get, the real web pages those answers were written from, where in the buying journey you lose them, and a specific list of fixes anchored to that evidence, plus one re-run to measure whether the fixes worked. One brand, about ten minutes, and every finding is yours to keep. No meter, no lock-in.
How it's built
The product is shaped by a few rules we won't bend, because a report you can't check is just another opinion.
We print the error bars, not a smooth line
AI answers vary run to run. Everyone knows it, nobody shows it. No single answer decides your score: scored questions are re-asked until a majority settles them, and you see the range, every single run is marked SNAPSHOT, and movement is only ever shown by re-asking the identical question set and comparing whole sets. When we say you moved, you moved. When nothing moved yet, your report says so instead of selling you noise.
Every number opens its receipt
Click any figure in a report and you see the stored raw answer, or the actual page, behind it. Nothing asks to be trusted; everything can be checked. Where we couldn't verify something, the report says "unverified". It never guesses.
We never invent a number
No made-up "AI search volume", no opaque one-to-a-hundred score, no llms.txt file sold as a fix. We measure what the engines actually do through their official APIs, and we tell you the limits of what we tested.
The full method (nine principles, each one checkable in your own report) is written up on the methodology page.
Saylent is open source (Apache-2.0). Run your own: github.com/yotambraun/saylent · docs