Case studies
Real work, real surfaces, every number on a receipt.
Most "AI visibility" is vibes. Ours is a measurement: citation assets we actually built and published, the engines they earn, and the before/after on the same locked battery — with the dated transcript behind every single answer.
How every case is measured
We measure first, then act
Before any work begins, we run a 400-datapoint baseline on your locked prompt battery. That's the world as it is today — not a guess, a measurement. Every claim we later make about movement is anchored to it.
Four surfaces, measured independently
ChatGPT Search, Perplexity, Google AI Overviews and Claude each read the web differently and cite different sources. We never collapse them into one number — you see the rate per surface, because that's where the truth is.
Repetition kills the noise
A single AI answer is a coin flip. We run every prompt five times per surface and read the rate across all of them, so a 'cited' result means a pattern, not a lucky draw.
Every number has a receipt
For each of the 400 runs, we store the full AI answer, timestamped. When your dashboard says 'cited 58%', you can open the transcripts and read the answers that earned it.
The report shape
Baseline versus endline, per surface — the picture you log in to.
This is the exact shape of the movement chart on your dashboard: citation rate per surface, before and after the 14-day cycle. The figures shown here are an illustrative sample — your numbers come from your own measured runs.
See it on a live demoSample figures to show the report shape — citation rate per surface, before and after. Not a measured client result. Your numbers come from your own 400-run baseline and endline.
What lands in your hands
The evidence pack — proof you can forward to your CMO.
Baseline + endline CSV
Every run — prompt, surface, rep, whether you were cited — for both measurements. 800 rows, nothing hidden.
Dated transcripts
The actual AI answer for every run, timestamped. The receipts behind every number on your dashboard.
One-page movement report
Baseline vs endline, per surface and per prompt, with the threshold verdict stated plainly.
Where the citations landed
The pages and placements the engines pulled you from — so you know exactly what moved the needle.
What we commit to
The measurement and the work — not a result we can't control. Here's exactly what that means.
We guarantee the measurement and the work — not a number we can't control.
- The measurement
- 20 co-written prompts × 5 runs × 4 surfaces = 400 datapoints, baseline and endline, with the full dated transcripts of every AI answer — so the before-and-after is a recorded fact, not a feeling.
- The work
- On-page rewrite to the citation spec, Wikidata entity scaffolding, and earned placement on a DR50+ listicle already ranking for your prompt-cluster — the three levers that actually move citations.
We can't promise a specific citation result — nobody honestly can. AI answers change constantly and aren't in anyone's control. We promise the measurement is real and the work is done.
Measured honestly
The whole point is a number you can trust. So the boundaries are hard:
- ✕Prompt injection, hidden text, fake citations, or anything that violates an LLM provider's content policy.
- ✕Promise you'll be cited by base-model ChatGPT in general — only by the surfaces that browse the live web.
- ✕Accept broad-head topics ('best CRM'). We ship long-tail prompts where movement is realistic.
- ✕Make you read a tracking dashboard and call it a service. We do the work and send the evidence.
Trust the number, not the pitch.
Register free and walk a live sample campaign — baseline, endline, transcripts and all — before you spend a cent.
Free to register · live demo dashboard · no card.