Being mentioned in an AI answer means you were present. Being recommended means the answer preferred you. They are different outcomes, and the gap between them is where most brands overrate their position. A mention feels like a win, but if the same answer recommends a competitor, you are the runner-up in the only sentence a buyer reads. Presence gets you into the answer; preference is what the answer actually does with you.
Presence and preference are two questions
Presence asks: did the answer name you at all? Preference asks: did it put you forward as the choice? An answer can name five brands and recommend one. Four of them are present; one is preferred. If you measure only whether your name appears, all five look equal, and they are not.
Buyers do not weigh a list of names evenly. They act on the recommendation, the first pick, the 'if you want X, go with Y'. Presence without preference is being in the room while someone else gets hired.
The states between absent and recommended
It helps to name the ladder. Absent: the answer does not mention you. Unlinked: it covers your category but never names you. Linked: it names you as one option. Recommended: it puts you forward as the pick. Misrepresented: it names you but gets something wrong. Each state is a different problem with a different fix.
Rolling every tracked question up to one of these states, per engine, turns a vague sense of how you are doing into a precise worklist. Absent means earn a mention. Linked means earn the preference. Misrepresented means fix the facts. You cannot act on 'we are visible'; you can act on 'we are linked but not recommended on eight of our top questions'.
Why the distinction changes what you do
If you treat presence as the goal, you optimize for being mentioned anywhere, which is easy and low-value. If you treat preference as the goal, you optimize for the reasons an engine recommends: clear positioning, credible third-party sources, accurate specifics, and content that answers the exact question better than the alternatives.
Preference is earned through the sources the model trusts and the clarity of your case, not the volume of mentions. The move from linked to recommended usually comes from stronger evidence: a comparison page that holds up, reviews that say the right thing, a source the engine already cites getting your facts right.
Measure preference, not just presence
A presence metric counts mentions. A preference metric weights them: were you recommended, named first, or described as the best fit, versus merely listed. AI Presence should be weighted by prominence, not a flat count, so a recommendation counts for more than a passing name-drop.
Read the two together. Rising presence with flat preference means you are getting into more answers without winning them, which is a content-and-sources problem, not an awareness one. That is the signal to stop celebrating mentions and start earning the recommendation.