The terms, explained plainly.

The category Caelis belongs to: one system that finds the demand, builds the answer and measures whether it worked.

Most tools do one thing: measure visibility, suggest keywords, or produce content. An operating system for AI search connects the whole chain — discover demand, understand the business, build what's missing, publish it, measure the outcome and improve it — so no part of the work stalls because a human has to carry it to the next step by hand. When we say “Caelis” without further qualification, this is what we mean.

The picture of where the business is visible or absent in AI answers to questions Caelis has written about the market.

Caelis tests questions about your market against the AI models and measures the answers — who gets mentioned, who gets recommended instead, and which sources are used. Caelis writes the questions; the answers are real and measured. Where search data exists, Caelis also shows what people actually search for. It's not a general market analysis — and it widens as more dimensions (market, language, season) are actually measured on an ongoing basis.

Connects what the market is asking for with what the business can offer — not a chatbot you talk to.

The Agent is what actually executes: it connects the demand in the Demand Landscape with your services, knowledge and positioning, prioritizes the opportunities and puts Service Engines to work. Its primary interface isn't a chat window, because its job isn't to answer you — it's to do the work and put it in front of you for approval.

The knowledge base about the business — services, customers, markets and approved, sourced facts.

Previously called Business Brain — now part of Caelis Agent. It's what the Agent knows about your business, and the context behind every recommendation and every Living Asset it builds. It's no longer its own step in how Caelis works externally — it's the knowledge the Agent uses to connect demand with what you can actually offer.

A digital asset your business owns, which is measured and improved after publication.

An ordinary page is published and then sits there. A Living Asset is published and then followed: measured against how AI answers the questions in the market, against who gets recommended instead, and against what it returns in enquiries or sales. If the questions or the competition change, Caelis proposes changes to the page that already exists rather than making a new one. Examples: pricing guides, comparison pages, service pages, calculators.

The step where Caelis Agent decides what to do first.

Caelis finds more opportunities than there is capacity to act on at once. The agent ranks them by expected impact and picks the highest-priority one — the rest wait, they don't disappear. This step is what makes Caelis a system that acts, not just a list of what's missing.

The step where Caelis checks whether the work actually worked.

Caelis runs real, weekly measurements of how the business is mentioned, recommended and cited in ChatGPT, Claude, Gemini and Perplexity — real numbers, not assumptions. The measurement is what decides whether something succeeded, not a guess about it.

The step where the measurement becomes the next priority.

What didn't work becomes a new recommendation for Caelis Agent. Living Assets are improved rather than replaced, and the cycle starts again with what's being measured now.

One bounded job the system can perform, switched on or off per customer.

The engines are the building blocks of what Caelis does: one measures visibility across the AI models, one finds demand, one builds content, one publishes, one measures outcomes. They are installed per business, and an engine that isn't switched on can't do anything. It's also how new capabilities roll out: one customer first, not everyone at once.

Whether your business appears at all when someone asks an AI model about something you sell.

This isn't the same as ranking in Google. An AI model answers with one composed recommendation, not ten links — either you're in the answer or you aren't. Caelis measures this by asking the same questions weekly of ChatGPT, Claude, Gemini and Perplexity and recording who actually gets mentioned.

How large a share of all mentions in the AI answers are about you, compared with competitors.

When several businesses are mentioned in the same answer, Caelis calculates how large a share of the total mentions is yours. Your own share and each competitor's share are calculated from the same answers, so the numbers can be compared directly — it isn't an estimate. Measured over time, it shows whether you're gaining or losing ground, not just whether you're mentioned or not.

An AI model using a source in its answer and stating where the information came from.

Being mentioned and being cited are two different things. A model can recommend you without linking to anything, and it can cite a source without recommending it. Caelis separates the two, because they call for different work: being mentioned is about reputation and coverage, being cited is about there being a page that actually answers the question.

A question phrased the way a customer would ask an AI model.

Where SEO counts keywords, Caelis counts prompts — whole questions, phrased the way a customer would say them: “who delivers X in Bergen”, “what does Y cost”, “is A or B better for me”. A prompt Caelis monitors is asked again every week, so changes in the answer are measured rather than assumed.

Making content legible to AI models, the way SEO does for search engines.

The abbreviation stands for AI/Answer Generative Engine Optimization. In practice the difference from SEO is less dramatic than the acronym suggests: concrete numbers, named sources, clear structure and unambiguous facts help both. AGEO doesn't replace SEO — they pull in the same direction on the same pages.

The program an AI provider uses to read web pages.

Many firewalls and CDNs block these along with other automated traffic, without anyone having decided to. The result is a website invisible to exactly the models you want to be recommended by. Caelis checks access daily for every published asset.

Tying an enquiry or a sale back to what actually started the customer journey.

Caelis reports only what can be traced end to end: a visit to a published asset, a click onward through a link on your own domain, and a recorded outcome at the other end. If the chain can't be closed — because an AI app hides where the visit came from, for instance — it's reported as uncertain, not as a result.

How long it can take from first contact to a sale and still count. At Caelis: 180 days.

B2B purchases take time. A 30-day window would systematically hide the value of content that starts buying journeys long before the decision is made. 180 days is chosen because it mirrors how these purchases actually run, not how quickly the numbers look good.

Measurement that happens on your own domain, without third-party tracking.

Published assets measure their own traffic rather than loading tracking scripts from other companies. That gives data that doesn't disappear with ad blockers or changes in browser policy, and it keeps the privacy picture simpler: fewer parties learn who visited what.

A link on your own domain that records the click before sending the visitor onward.

Instead of putting tracking parameters in the URL — which frequently get stripped along the way — the click goes via a short address on a subdomain you own, for example go.yourcompany.com. The source is therefore recorded before the visitor even lands on your website.

Nothing is published without a yes. Graduation is how one type of work gradually earns the right to run on its own.

Everything Caelis proposes is logged and waits for approval from someone on your side. If you've approved the same type of work ten times running without rejecting any of it, that one type can stop asking every time. It applies only to work types you switched on yourself, and only to that one type — nothing else.

Payment calculated from documented value. Not the default at Caelis.

All standard plans are flat-priced. In select Enterprise agreements, a fee tied to traceable, documented value can be agreed in addition to the flat price — never instead of it, and never on value that can't be traced.