How it works

Learn once. Apply everywhere.

Jinn learned your brand first.

Every Jinn product starts from the same place: a working model of your brand, extracted from what you have actually made. This page shows the mechanism.

01 It learns

It learns.

Jinn reads your site, your products, and what you have published, and assembles 346 brand signals into one working record - inspectable, field by field.

Open the groups. Every signal is a real, named thing Jinn extracted - none of it is a prompt.

  • #602F00
  • #8a4a10
  • #c98528
  • #f2e6d6

Example record · not a customer

The schema underneath

What Jinn goes looking for when it reads a brand: six field groups, 346 brand signals, every one a named field.

  • Identity42

    • name
    • category
    • founding story
    • positioning
  • Palette28

    • primary
    • secondary
    • accent
    • usage rules
  • Voice71

    • cadence
    • convictions
    • banned words
    • register
  • Evidence96

    • claims
    • sources
    • proof points
    • receipts
  • Audience54

    • who
    • jobs-to-be-done
    • objections
  • Competitors55

    • set
    • gaps
    • differentiation lines

The full record breakdown

What the machine sees

A raw slice of the record above - the field values Jinn extracted, as the pipeline reads them.

identity
ancestral whole-food protein · Explorer archetype
voice
direct · minimalist · all cow, no bull
voice.banned
"proprietary blend" · "guilt-free"
evidence
founder rebuilt his body with real food after a life-altering accident
audience
carnivore devotees · gut-conscious achievers · 18-65
competitors
mass-market whey benchmarks · organ capsule leaders
palette
#602F00 · #8a4a10 · #c98528 · #f2e6d6

EXAMPLE · PALEO PRO DNA · NOT A CUSTOMER

Generic run · no DNA

02 It wears

It wears you.

The same brief, run twice. The generic run is what any model writes by default. The branded run was written from the record above - the orb re-authors it in front of you.

Which signal produced which output

Nothing here is a style filter. Each line of the branded run traces to a named signal in the record.

  • SignalVocabulary banned list"proprietary blend"Never says "proprietary blend"
  • SignalSlang policyall cow, no bull"All cow, no bull"
  • SignalPositioning wedgereal food, real fuel · four ingredients per scoop"Four ingredients per scoop"
  • SignalVocabulary banned list"guilt-free""Guilt-free" appears nowhere - a check, not a suggestion
  • SignalTonal attributesdirect · minimalistEight words, two sentences, zero hedges

EXAMPLE · PALEO PRO DNA · NOT A CUSTOMER

03 It works

It works.

A product running on the worn brand. This is a Ghost draft composing in the voice the record defines - banned words enforced, not suggested.

Ghost · composeDraft in voice

Never saysproprietary blendguilt-free

The pipeline behind the draft

Nothing here is a filter on top of a generic draft. The record runs end to end.

  1. 01 record

    The worn record rides in: 346 brand signals assembled into the prompt.

  2. 02 compose

    The draft is written against cadence, convictions, and the banned list.

  3. 03 checks

    Every post passes voice checks before it publishes.

  4. 04 publish

    End to end, all four platforms.

    • X
    • LinkedIn
    • Bluesky
    • Facebook

Analytics is X-only today.

The same record feeds every hand: written posts, on-brand ads, visibility audits, strategy with receipts.

Models & infrastructure

You never pick a model.

Every job in the suite is routed to the class of model it needs. The table that decides is one file, kept current as vendors ship new families - you do not have to figure it out, keep up, or experiment. It just works.

The routing table

Five workloads, each with its own rule. Overrides are allowed, but logged with a reason.

  • Visibility checks, run constantlymonitoringFast, small modelsFrontier models are forbidden on this path - at scale the cost ceiling breaks immediately. A premium plan buys more samples and more engines, never a bigger monitoring model.
  • Judging what an engine got rightfidelityMid-weight · frontier at premiumAccuracy adjudication runs a mid-weight model; premium tiers escalate to frontier.
  • High-volume short-form copypin_copyFast, small models · every tierA batch of 75 pins times several calls each cannot afford a frontier model, and short brand-voiced copy does not need one.
  • Long-form strategy documentsdeliverableWorkhorse to frontier, by roleNarrative sections run the workhorse class many times per document; synthesis-heavy documents and the point-of-view letter escalate to the most capable model available. On an upstream error the call falls one tier with a logged warning - never a silent substitution.
  • Strategy briefsbriefAlways frontierRegardless of tier - opt-in and low volume, so the ceiling holds.

When a vendor ships a new model family, the swap is one table entry - shipped to every product at once.

One gateway, every call

Every model call in the suite passes through a single gateway. What that buys:

metered
Tokens and cost are recorded on every response - every generation is accounted for.
ceilinged
Spend is reserved before the call and settled to the actual cost after. If the meter is unreachable, the call is refused - fail closed, never an untracked spend.
bounded
Every call carries a deadline and is aborted past it - a hung engine never hangs the product.
typed
Failures come back named - rate limit, refusal, timeout - so products handle them instead of guessing.
priced
Four engine families priced through one shared rate lookup: Anthropic, OpenAI, Google, Perplexity. An unknown model id is costed at a deliberately expensive fallback and flagged - never silently under-counted.

Why this matters now

Search is being re-priced from rankings to citations.

The old game was ranking on a page of blue links. The new one is being the answer an AI engine cites when a buyer asks. That traffic is still a sliver of the total, but it converts far harder than a search click, and it is concentrated in a handful of engines. The brand the engines describe correctly wins the citation. Everything above is how Jinn earns it.

What the industry is seeing

  • ~1%of site traffic arrives from AI answers today - still a sliver
  • ~4.4xhigher conversion than an organic search click, when it does arrive
  • up to 23xthe conversion lift for B2B SaaS specifically
  • ~92%of that AI-referred traffic comes from ChatGPT alone

Third-party industry figures, not Jinn measurements. Volume, conversion, and share estimates from independent 2025-2026 analyses (Lantern, Pixis, Search Engine Land), compiled in our channel-evidence research. Directional, and every number carries the range shown.

Runs on the Jinn Brain.

Every Jinn product starts from your Brand DNA: your voice, visuals, products, and customers, extracted from what you have actually made. Learn once. Apply everywhere.

What we measure, and how

Stop guessing. Start knowing.

Jinn