What is Cumae?
The short definition
Cumae is a business-controlled information layer for AI-driven discovery. Instead of leaving AI systems to assemble an answer about your business from whatever they can find, you maintain one approved record and decide who is allowed to read which parts of it.
A Cumae record has three jobs:
- Publish. Establish an identity record with your company name, canonical website, locations, offerings and a path to current authoritative information — then build the deeper record behind it.
- Monitor (in development). Test a defined library of material business questions across designated AI services, compare the answers with your approved record, and document material inaccuracies with evidence.
- Govern. Give registered applications, individually approved partners and your own company-controlled agents different levels of access. See How Cumae Access Policies Work.
Why a record and not a webpage
Your business information already exists — on your website, in directories, on marketplaces, in reviews. The problem is not that it is missing. The problem is that none of it was written for a machine to reason over.
A web menu tells a person what a sandwich costs. It does not tell an AI agent what is in it, what it weighs, which allergens are present, which are absent, and which are absent but still at cross-contact risk from shared equipment. That third state is exactly what a customer with a serious allergy is asking about, and almost no menu carries it.
A Cumae record does. It is structured as addressable blocks, each with its own identifier, source and freshness stamp, so an agent can retrieve one fact without parsing a page. That is the difference between content a machine can scrape and information a machine can rely on.
Precision includes saying no
The most useful thing in a Cumae record is often the limit. These lines are taken verbatim from a live demonstration record:
“Wheat flour is airborne in the same room and the press is shared, so Copperline describes it as a gluten-free roll, never as a gluten-free sandwich and never as coeliac-safe.”
“CRSTBL publishes what the source of truth states; it does not compute or certify them. If a figure matters medically, confirm it with the counter before ordering.”
“There is no checkout here and no card is taken… Anything claiming to have completed a Copperline payment did not.”
An AI system reasoning over a record like this cannot accidentally invent a safety assurance the business never made. That is what makes a record trustworthy — not that it says more, but that it is precise about what it does and does not cover.
What about AEO and GEO?
AEO (AI Engine Optimization) and GEO (Generative Engine Optimization) describe the practice of making business information discoverable and usable by AI systems. They are the AI-era counterparts to SEO.
Cumae is how you do that work with a source you control rather than by guessing at someone else's ranking model. But it is worth being precise about the mechanism: publishing a record does not push information into any AI system. It gives search engines, AI crawlers and registered applications something authoritative to retrieve, on their own schedules and under the access policy you set.
Cumae and your Google Business Profile
Keep it. A Google Business Profile gets your business into Google Maps and Search, and it does that job well. It was built for a world where customers searched by keyword and clicked a link.
Cumae is built for a world where customers ask a question with conditions attached — and where a directory listing has nowhere to put sub-ingredient declarations, allergen states, warranty processes, or partner-specific terms. The two complement each other rather than competing.
Cumae and CRSTA: different jobs
They are separate products, sold separately, and either works on its own.
- Cumae publishes and governs the authoritative business information used in AI-driven discovery — the source behind the answer given about you.
- CRSTA applies business knowledge in conversations on your own surfaces — the answer you give, on your website, kiosk or app.
See How CRSTA and Cumae Work Together and What is CRSTA?
What Cumae does not do
A few honest limits worth naming:
- It does not control what any AI system says. Cumae governs the information you publish and the access policies you configure. CRSTBL does not control third-party AI platforms and cannot guarantee how any of them answer.
- It does not publish everywhere. Registered and allowlisted applications retrieve the current approved version under their integration terms. Open web and external AI systems refresh on their own schedules.
- Monitoring is bounded. It applies to defined questions and designated AI services, limited by plan, question library, platform, location, language and testing frequency. It is not an unlimited review of every possible AI-generated statement, and it does not monitor every menu item, SKU, price or real-time inventory level.
- Correction is best-efforts. CRSTBL performs defined correction and escalation activities through available platform and source channels, and cannot guarantee that a platform will change a particular answer, citation, ranking or recommendation. What CRSTBL can change is what those systems retrieve.
Frequently asked questions
Does Cumae replace my website?
No. A Cumae record carries depth a marketing site was never built to hold, and it is designed to be read by machines as well as people. Your website keeps doing its job; the record answers the questions the website cannot.
Do I need CRSTA to use Cumae?
No. They are priced separately and sold independently, and Cumae plans are never blended with CRSTA conversation allowances. Start with either.
Who decides what is in the record?
You do. Only information your business has reviewed and approved is published or made available through the access policies you configure.
Can a platform partner access everything?
No. What an application can retrieve is set by the access policy each participating business configures — Open, Registered, Allowlisted or Private. Private information is not available to external platform partners at all. See Platform Partners.
Is there a free option?
Yes. Cumae Publish is free and publishes one self-asserted public identity record with Open access — no domain or entity verification, and the record declares that plainly. It does not include the full record, monitoring, incident alerts or partner controls. A paid Cumae Record adds verification and Registered access, and lets you configure Allowlisted and Private views. Ownership of a Publish record can be claimed or disputed through CRSTBL’s claim process. See Cumae pricing.
See a live Cumae record.
The fastest way to understand a Cumae record is to read one. Copperline Sandwich Co. is a demonstration record built by CRSTBL — the business is fictitious, the structure and depth are real.