Jinn assembles 346 brand signals into six field groups. This page is the breakdown: what each group holds, what extraction reads to fill it, and what the machine actually sees.
A Brand DNA record is extracted from what a brand has actually made - its site, its products, its published work. None of it is a prompt.
What it is
A Brand DNA record is one structured record of 346 brand signals, organized into six field groups, extracted from what a brand has actually published.
Where it comes from
Extraction reads the brand’s site, products, and published work; every signal is a named field with a recorded value, not a prompt.
What it feeds
Every Jinn product reads the same record: written posts, on-brand ads, visibility audits, strategy with receipts.
The counts below are the record’s own group sizes, and the named signals in each group are the public projection of that group’s fields. Together the six groups hold the 346 brand signals.
Identity42 signals
Identity signals record who the brand is: its name, category, founding story, and positioning. Extraction reads the about page, the origin story, and how the brand introduces itself when nobody is watching. This group anchors every other one - a voice without an identity is a costume.
Palette28 signals
Palette signals record the visual system as usable rules: primary, secondary, and accent colors plus their usage rules. Extraction reads the rendered site and published assets, then stores colors as named slots with rules attached, not a mood board. That is why generated work can hold a hex, not a vibe.
Voice71 signals
Voice signals record how the brand talks and how it refuses to talk: cadence, convictions, banned words, register. Extraction reads published copy at every register - product pages, posts, packaging lines - and stores the patterns as enforceable fields. Banned words are enforced downstream, not suggested.
Evidence96 signals
Evidence is the largest group on purpose: claims, sources, proof points, receipts. Extraction reads what the brand asserts about itself and what actually backs it, so anything generated later can cite a recorded fact instead of inventing one. This group is why outputs carry receipts.
Audience54 signals
Audience signals record who the brand is for: who they are, their jobs-to-be-done, and their objections. Extraction reads reviews, testimonials, and the language real customers get answered in, and stores buyers as named segments rather than one imagined persona.
Competitors55 signals
Competitor signals record the market the brand sits in: the set, the gaps, the differentiation lines. Extraction reads the category around the brand and stores where the brand genuinely differs, so positioning claims have a recorded edge to stand on instead of a hunch.
Group sizes are read from the record taxonomy itself, so the numbers on this page cannot drift from the product.
A raw slice of a real extracted record - the field values as the pipeline reads them. This is the worked example used across the how-it-works pages, not a customer.
EXAMPLE · PALEO PRO DNA · NOT A CUSTOMER
That is one worked example. Jinn builds the same record from your brand. See your own record
Nothing generated from the record is a style filter. When a Jinn product writes a line, picks a color, or refuses a word, that behavior traces to a named signal: a banned word in the voice group blocks a phrase, a positioning signal in the identity group shapes a claim, a recorded proof point in the evidence group backs it.
That is the point of holding signals as named fields instead of a prompt: the work is inspectable in both directions. You can open the record and see what Jinn learned, and you can take a finished output and ask which signal produced which line. The full tour shows this trace running live.
The names on this page are the public projection. The full field list is the product.
Each group above shows its real, named signals - the same ones published across these pages - and its true size. What stays private is the rest of the field list and the extraction prompts that fill it.
The overview is the tour; this page is the reference for one layer. The other layers get the same treatment as they land.
The record on this page is built by reading a real brand. Run the free read and see what Jinn extracts from yours.