> ## Documentation Index
> Fetch the complete documentation index at: https://docs.get3rd.com/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Ontology

> A fixed methodology that asks the AI models a purpose-built set of questions to show how they see, understand, trust, and recommend your brand.

**AI Ontology** runs a fixed methodology against the AI models you track: a purpose-built set of questions, put to every model several times over, designed to show how they see, understand, trust, and recommend your brand.

The result is one wheel — sixteen signals across four quadrants.

<Info>
  AI Ontology is in **beta** and switched on per organisation. If you see a request-access card, the feature is on your plan but not yet enabled for you.
</Info>

## What it measures

Being visible and being chosen are different problems. A brand can be named constantly and never recommended; another can be understood perfectly and never surface. The four quadrants keep those problems apart, and each is made up of four spokes.

<Tabs>
  <Tab title="Seen">
    **Does AI surface you at all?** Your existing visibility, snapshotted alongside the interview.

    * **Prompt coverage** — how many of your prompts you appear in at all
    * **Share of answer** — how often you appear across the runs measured
    * **Prominence** — how high you rank in the answers you do appear in
    * **Market consistency** — your weakest market's visibility as a share of your strongest
  </Tab>

  <Tab title="Understood">
    **When AI talks about you, does it get you right?** Graded against the positioning you set in your brand profile.

    * **Entity clarity** — whether the models resolve your bare name to you, not a similarly named company
    * **Offering understanding** — whether they know the products and services you actually sell
    * **Positioning fidelity** — whether they describe you the way you want to be described
    * **Claim endorsement** — whether they'd back your own words about yourself, and can point to something specific when they do
  </Tab>

  <Tab title="Trusted">
    **How does AI rate you against your competitors?** One ranked comparison per value, across your competitive field.

    * **Products & services** — the quality and value of what you sell
    * **Innovation** — how quickly you adapt and how far ahead you're seen to be
    * **Integrity & responsibility** — ethics, openness, and responsibility to society and the environment
    * **Leadership & strength** — leadership, financial strength, long-term prospects, and being a place people want to work
  </Tab>

  <Tab title="Recommended">
    **Would AI actually put you forward?** Shortlists, forced choices, and head-to-head challenges.

    * **Shortlist inclusion** — whether you make a spontaneous shortlist at all
    * **Shortlist prominence** — when you do make it, how near the top you land
    * **Competitive preference** — whether the models pick you when forced to recommend exactly one of you and your competitors
    * **Conviction** — whether they keep a buyer with you when a switch is proposed, and move a buyer to you when it runs the other way
  </Tab>
</Tabs>

## Tell 3RD how you want to be understood

You set out how you want AI to understand your brand in your **Brand profile**: the positioning you want to hold, the attributes you want tied to your name, and the ways you don't want to be described.

AI Ontology then rates whether the models actually see you that way. That comparison is the Understood quadrant — without it, there is nothing to grade the models against.

## What you get out of it

A diagnostic you run when you need to understand your position, not a number to track over time.

* **The delta between how you want to be perceived and how you actually are.** Your profile states the positioning and attributes you want; the wheel shows what the models really hold. The distance between the two is the work.
* **Whether you stack up against your competitors.** The same questions put the whole field side by side, so you can see where a rival is rated above you and by how much.
* **The reasons behind the choice.** When the models are forced to pick one of you, they say why — what counted for you, what counted against you, and what won it for the competitor. That reasoning is the most actionable thing on the page.

Together they tell you where you're weak and what would have to change for AI to choose you.

## How to use it

* **A claim you make that no model repeats, backed by nothing but your own site.** You don't need more copy on your own pages — you need someone else to say it. See [Authority and outreach](/drivers/authority-and-outreach).
* **A category or association you want to own that never comes up.** There's nothing for the models to read. That's a content gap — see [Content](/drivers/content).
* **Models that describe you accurately but still pick a competitor.** Not a visibility problem. Their stated reasons tell you which proof points your value proposition is missing.
* **An area where the models say nothing at all.** Your brand isn't legible there yet. Start with the basics — being identifiable in that area at all — before optimising anything.
* **Your name resolving to a different company.** Fix this before anything else. Every other signal is measuring somebody else.

## Related

* [Brand profile](/workspace/brand-profile) — where your ontology lives
* [Visibility](/metrics/visibility) — the live metric behind the Seen quadrant
* [Identifying your competitors](/setup/competitors) — the field Trusted and Recommended measure you against
* [Playbooks](/actions/playbooks) — turning these gaps into prioritized work