One question. Five engines. Fifteen answers. One honest number.
The clearest way to explain how Treyci measures AI visibility is to follow a single buyer question all the way through.
From buyer question to board number, in six steps.
The example uses an invented security brand, Northwind. Every real Treyci measurement follows exactly the same steps, for every question in your set and for every competitor.
Start with what buyers actually ask.
We build your question set from your real category: full questions phrased the way people talk to AI, not keywords. Best-ofs, comparisons, alternatives, pricing, reviews.
25–250 questions depending on plan. On Growth and up, you can add your own.
Ask every engine separately.
The same question goes to ChatGPT, Gemini, Perplexity, Grok and Claude through their official APIs, with web search on. Each engine builds its answer from different sources, so each one gets its own score.
Starter covers ChatGPT, Gemini and Perplexity. Model versions are pinned, so your score never moves because of an unannounced model change.
Ask again. And again.
AI answers change from run to run. A tracker that asks once is reporting luck. Treyci asks every question three times per engine, every month, and reports the pattern across all of them.
One question becomes 15 answers. Only Perplexity gave the same verdict all three times. That agreement rate becomes your consistency score.
| Engine | Run 1 | Run 2 | Run 3 | |
|---|---|---|---|---|
| ChatGPT | Rec | Rec | Named | |
| Gemini | Named | — | Named | |
| Perplexity | Rec | Rec | Rec | 3/3 ✓ |
| Grok | Rec | Named | Rec | |
| Claude | — | Named | — |
Score it with a public formula.
An AI model reads every answer and records, for each brand, whether it was named and whether it was recommended. Whether your site was cited is read directly from the answer’s source links, with no judgment involved.
The rates feed one formula, the same for every competitor: 50% recommended, 30% named, 20% cited.
| Recommended 7 of 15 | 46.7% | × 0.50 | 23.3 |
| Named 12 of 15 | 80.0% | × 0.30 | 24.0 |
| Cited 3 of 15 | 20.0% | × 0.20 | 4.0 |
| Score for this question | 51.3 |
Repeat for every question and every competitor.
Do this across your whole question set and you get one visibility score, a rank against your competitors, and a breakdown by engine. All of it lands in a monthly brief: what moved, why, and what to do next.
If an engine fails during a run, its missing answers are re-run before scoring, so a gap never shows up as a drop.
Visibility rose 3.4 points, led by Perplexity after the new comparison pages were cited in 9 answers.
Vantage still owns “best EDR for mid-market”. Gemini recommended them in all three runs.
Next: earn a verdict on the two review sites Gemini cites most.
Click any number back to its source.
Every answer is stored word for word, with the sites it cited. Any number in Treyci clicks through to the sentence the AI wrote. It’s a record of what AI said, not our opinion of what it thinks.
g2.comnorthwind.com — citedvantage.com/compare
Three decisions that make the number trustworthy.
Repeat runs over daily snapshots
A monthly, three-run measurement is more reliable than a daily single answer, at a fraction of the cost. We don’t sell “real-time”, because day-to-day swings in AI answers are mostly noise.
Every engine on its own
One blended “AI score” hides the engine you’re losing on. You see each engine separately, and the sources each one trusts.
Receipts for every number
A score you can’t check is a guess. Every Treyci number links back to the answers that produced it.
What this measurement can’t tell you.
Personalized answers
Consumer apps tailor answers using memory, chat history and location. API answers are a clean baseline, not every answer every user sees.
Guaranteed rankings
The engines are third-party systems. We measure what moved and why; nobody can promise a position.
The full picture, from the free check
The free check is a quick snapshot: 12 questions, ChatGPT and Gemini, one run each. The full method starts with your baseline.
Follow your own question through the engines.
Run the free check and see where your brand lands in the answers buyers are getting today.
Built by Helix Apps — the methodology behind the Helix AI Visibility Audit, productized.