Ask any frontier model to compare you to your three closest competitors. Whatever comes back is now a meaningful share of your brand, because a growing number of buyers form their first impression of you in exactly that way, without ever loading your homepage. Almost nobody is auditing that impression, and the teams who are auditing it are finding descriptions of themselves they would never have approved.
This is a strange kind of brand surface. It is not owned media and it is not earned media. It is a compression of everything written about you, weighted in ways you cannot see, delivered with more authority than either. A buyer who reads a paragraph about you from a model treats it roughly the way they used to treat an analyst note, and it costs them four seconds.
Your positioning is no longer what you publish. It is what gets summarized back.
The practical implications are less exotic than people expect. Models summarize consensus, so the lever is the corpus. Clear, consistent, unhedged language across your own properties. Third party writing that uses the same terms. Documentation and comparison pages that state plainly what you do and do not do, because ambiguity gets resolved by the model in whatever direction the rest of the internet leans, which is usually toward your largest competitor.
There is a harder implication too. If the summary of you is wrong, the fix is not a campaign. It is publishing enough unambiguous material that the wrong version stops being the consensus, which takes quarters and does not feel like marketing while you are doing it. What I keep wanting is a way to measure it, because right now every conversation about this happens from impression. Someone runs one prompt, screenshots the answer, and the room reacts. That is not evidence. Something that queried the major models on a schedule, across a whole competitive set, would turn positioning back into an argument about facts. I have not found that tool. I am increasingly convinced I am going to have to build it.