Case study

Citelight

A SaaS visibility product, zero to live. When buyers ask AI who to trust, Citelight audits whether you are in the answer, platform by platform.

Own venture·2026·Live at citelight.ai
The surfaces

The audit is the product.

citelight.ai · visibility audit
CitelightVisibility auditDemo data
ChatGPTCited
PerplexityAbsent
GeminiCited
Google AI OverviewsAbsent
2 of 4engines mention this agent when buyers ask. The audit shows exactly where, and where not.

Four engines, cited or absent, one honest score.

what the machines answer
who is the best real estate agent in fort lauderdale?
Answer · assistant
1.Harbor & Main RealtyWaterfront specialists, strong recent coverage.
2.The Reyes GroupYou.High volume in Victoria Park and Rio Vista.
3.Lindqvist ResidentialBoutique, well reviewed for new construction.

The moment the product exists for: a buyer asks, and the second name is you. Demo data.

The system

Its own tokens, nothing borrowed.

Measure what the machines say about you.
Söhne over a system grotesk stack. One weight family, one gold, nothing decorative.
#0A0A0BGround
#111114Surface
#26262BLine
#E8E8EBText
#E5B53EAccent
Voice
Measure what the machines say about you.
Accent
One gold. Everything else stays out of its way.
What shipped

A product, not a promise.

Product
A visibility audit across ChatGPT, Perplexity, Gemini, and Google AI Overviews, with per-platform results an agent can act on.
Marketing site
Live at citelight.ai. The audit reveal flow was designed and rebuilt until the result, not the mechanism, was the hero.
Brand
A cool, dark world with a single gold accent. None of it is shared with any other project I run.
Production
Deployed, monitored, and hotfixed in production. This is working software, not a deck.
The system behind it

One person, end to end.

This is the SaaS category of the practice: strategy, brand, product design, engineering, and deploy from one pair of hands, with AI as the multiplier. Citelight is also my own venture, which means the same system that clients hire runs my own money first.

What I would build next

Named openly, because unfinished is honest.

Claims audit
Per-platform result copy is paused until a veracity pass confirms exactly what the data supports. The product will only say what it can prove.
Visibility over time
Tracked runs, so an agent sees a citation trend instead of a snapshot.
The action layer
From where you are missing to what to publish next.