Methodology

Instrumentation, not opinion

Most "AI SEO" advice is somebody's screenshot. Treyci is a measurement instrument: the same questions, the same engines, repeated runs, every answer recorded — so the number you see in month six is comparable to the number you saw in month one.

Your category
real buying-intent prompts
4 engines
ChatGPT · Perplexity · Gemini · Grok
× 3 runs
every prompt, three times
Parsed & scored
every brand, every answer
The brief
what changed, why, what to do

Real buyer questions, not keywords

We build your prompt set from your actual category — full questions phrased the way people talk to AI, because that's what your buyers type. Standard sets run 25–250 prompts depending on plan; you can add your own on Growth and above.

best project management software for agenciesbest of
Asana vs Monday for a 200-person companycomparison
cheaper alternatives to Salesforcealternatives
how much does HubSpot actually costpricing

Four engines, because each is a separate battle

ChatGPT, Perplexity, Gemini, and Grok each assemble answers from a different graph of trusted sources. Brands routinely dominate one engine and vanish on another — and the engine where you're weakest is usually the one with the most buyers.

ChatGPT
the biggest buyer audience
Perplexity
the most transparent citations
Gemini
feeds Google's AI surfaces
Grok
the most brand-dense answers

Example mention rates for one real brand, one run — a 38-point spread across engines. Each column is a separate battle.

Three runs per prompt — the honesty mechanism

AI answers are stochastic: the same question can return different brand lists run to run. Single-run trackers (most of the market) report that noise as news. Every Treyci prompt runs three separate times each month, and results are aggregated:

“best work management platform for mid-size teams”
Run 1RECOMMENDED
Run 2MENTIONED
Run 3RECOMMENDED
=
R 2/3
reproducible measurement, not a lucky answer

We also publish your consistency score — the share of prompts where all three runs agreed — because rising consistency means the engines have settled on their sources, and that's a signal in itself.

Three levels of scoring, one score

Every answer is parsed for every tracked brand and coded on three levels, weighted the way buying actually works:

Recommended50%

Presented as a top option — the level that builds shortlists

Mentioned30%

Named anywhere in the answer — table stakes

Cited20%

Your own domain in the answer's sources — the trust signal

■ Recommended 50■ Mentioned 30■ Cited 20= Visibility Score / 100

Every number traceable to a verbatim answer

We store the engines' actual words and their citation lists. Any stat in Treyci can be clicked down to the sentence the AI wrote and the sources it used. Nothing in your dashboard is our opinion of what AI thinks — it's a record of what AI said.

Perplexity · run 2 · “best work management platform…”

…for teams that value visual planning and easy adoption, monday.com is the strongest choice, while Asana suits process-heavy organizations…

Sources: g2.com · techradar.com · monday.com

Monthly cadence, honestly labeled

Treyci runs monthly, on schedule, and says so. We don't sell "real-time" AI tracking because the underlying answers don't change in real time in ways that matter — and because a monthly 3×-repeat measurement beats a daily single shot on every statistical axis while costing you a fraction as much.

See it against your own brand.

Your baseline run answers the only question that matters: are you in the answers, or is your competitor?

Get your baseline

Built by Helix Apps — the methodology behind the Helix AI Visibility Audit, productized.