Skip to main content
Brand Studio is in beta. Your account needs the Brand Studio beta to see it.
Brand Studio is an advertiser’s Brand section. It shows how the advertiser’s brand shows up where people ask AI about it: today that is Nara, where a brand can have a Mini that answers when someone’s own Mini asks. Open an advertiser in the Apostra web app and choose Brand, or read the same facts through the V3 API.

What it shows

What you will never see

Nara sends counts, amounts and dates only. You never see the words of any talk, names, who asked, or which fans. A count of talks, offers, fans or sponsored results under 5 is withheld, and Brand Studio shows it as “fewer than 5”, never as 0. A sponsored option’s talks started and orders are also withheld when they would reveal individual orders. Organic orders are the exception: Nara lists them one by one, so they are counted exactly.

Simulated and stand-in parts

While Brand Studio is in beta, some parts are labelled Simulated or Stand-in:
  • A fictional brand’s results are simulated history. Nara never reports a brand on an .example domain, so Acme Outdoor and Fernhollow Audio always show simulated results.
  • Places at national scale and “asked for the nearest” counts are simulated: Nara does not report them yet.
  • Insights are simulated until Nara offers them.
  • The Founding Brand offer is a stand-in until it is a product you can buy.
A part without a label is real.

Who sees what

A brand is its own record in Apostra, outside any advertiser, so Coca-Cola is one brand whether Coca-Cola, its agency or both run it. Who you are to a brand decides what you see: Organic results are the brand’s measurement, so an operator entry grants them only when its scopes include measurement or all, or name no scopes, which AdCP reads as broad authorisation. An entry scoped to other activities alone, such as rights_clearance, reads the brand as anyone else does. An authorized_operators entry grants access only when it matches AdCP’s brand.json schema exactly. An entry with no brands, a brand id, country or scope the schema does not allow, or a date that is not an ISO 8601 timestamp such as 2026-01-01T00:00:00Z grants nothing. Organic results are not split by country yet, so an authorized_operators entry limited to some countries does not grant them. Paid results are different: they belong to the advertiser that paid. You always see what your own buys did, and never another operator’s spend on the same brand.

Reading it through the API

Brand Studio reads the V3 surface an agent reads. A brand’s id is its domain, or domain/brand_id when the domain’s brand.json lists several brands. The domain is lowercase and has no www.. You can pass it either way, for example www.Coca-Cola.com, and the read answers with the brand’s id, coca-cola.com. Read the brand with get:
  • The brand itself carries its name, url and logos from its brand.json, its house when it belongs to one, and your access: owner, operator or public.
  • presence returns registry (the brand’s AdCP registry record), trust (its trust.json verdict) and surfaces: Nara’s state (verified, verified_demo, not_verified, simulated or not_present) and whether the ChatGPT export is ready.
  • surfaceResults returns the brand’s organic results on each surface for the last 28 days, each count null when Nara withheld it, with a provenance of live or simulated. Only the owner and its authorised operators get it; anyone else finds it in unavailableIncludes with the reason.
  • Nara reports by domain, so a brand within a domain, such as pg.com/tide, has no presence or surfaceResults of its own: both come back in unavailableIncludes. Read the domain’s brand, pg.com, for them.
search with kind: "brand" lists the brands your advertisers name. Read what an advertiser’s own buys did with get on the advertiser and the paidResults include:
paidResults returns, per surface, its sponsored options and creator deals, with the same withheld counts and provenance. An advertiser’s brandIdentity.brandRef names its brand. Places are the advertiser’s catalogs of type: "store", read with search and get on kind: "catalog".