How it works
How it works
No dashboard scraping, no single-shot guesses. We query the engines directly, sample, and read the actual sources — then map the gaps to programs your data supports.
- 1
You tell us your brand
Brand, website, country, and optional competitors. Country sets the engine locale and the demand data.
- 2
We ask the engines and read the answers
We generate decision-stage buyer prompts, sample each across Perplexity, OpenAI and Claude, and read which brands and sources every answer cites.
- 3
You get a report you can act on
Per-brand visibility, the prompts competitors win, the source landscape, and a plan mapped to the programs your data supports.
The methodology, in full
No black box. Here’s exactly how the read is built.
Six steps, every one designed so you can check our work.
- 01
We generate the prompts your buyers actually ask
No taxonomy, no keyword list. For your category we generate decision-stage buyer questions such as 'best X for…', 'X vs Y' and 'alternatives to X', grounded in real buyer language. Your brand name never appears in the prompt, so the engines answer honestly.
- 02
We ask the engines directly, and sample
We query Perplexity, OpenAI and Claude through their APIs, not a dashboard, and sample each prompt several times because AI answers aren't deterministic. That's why we report 'N prompts, M AI responses analyzed,' not 'N prompts checked.'
- 03
We keep only the prompts that separate brands
A quick discovery probe runs every candidate once. We keep the prompts that actually separate winners from losers, plus 'open-territory' prompts where demand exists but no one is cited yet, and drop the flat ones that tell you nothing.
- 04
We read sources and presence, not vibes
For each kept prompt we read which brands are named, in what position, and exactly which sources the engine cited, typed by kind (Reddit, review site, blog, press). Presence is evidence-backed: a 'present' verdict carries the quote that justifies it.
- 05
We score, and we stay honest about confidence
We rank the prompts that matter and attach a confidence read. When our measurement and the consumer apps might diverge (most common outside the US), we say so, soften the language, and hold back priced recommendations until we've calibrated for your market.
- 06
We map the gaps to programs your data supports
Every gap maps to a specific program, recommended only when the source landscape supports it such as Reddit threads to Reddit Authority, or competitor blog content to Content/LLMO. The plan is priced, sequenced, and honest about what it can and can't do.
Why the careful wording
Why “across the engines we measured”?
We measure the AI engines’ web-grounded APIs, not the consumer apps. So every claim we make is phrased as “across the engines we measured (Perplexity, OpenAI, Claude), X appears in N of M responses”, never “ChatGPT recommends”. A claim a buyer can disprove in 30 seconds backfires; honesty is the strategy. When our measurement and the consumer apps might diverge (most acute outside the US), we soften the language and hold back priced recommendations until we’ve calibrated.
Your free report
See exactly how the read works on your brand.
Specific to your brand, across the engines we measure. Takes about 2-3 minutes, no card, no call required.