An AI brand content strategy built from the answers
What to publish and fix and in what order so that AI names you and describes you correctly. Built from the pages the models already read and measured after every change.
What you may be experiencing
Your content was written for a search box. It ranks and it still does not get you named. The agency brief says more thought leadership and nobody can say which answer it would change. One hero page carries the argument while the models sample everything you have ever published and repeat what they find most often. Your prices and product names are stated three different ways on your own site.
A content strategy for AI starts from what the models already read and not from a keyword list. We read every page behind the answers in your category and judge each one against the journey and the gap between those pages and yours is the plan.
The readings this page draws on are on the Features page under Site.
The three layers of the plan
A plan has to say what to be known for and where to be present and what each page needs to be read. We call it the three-layer content plan and each layer has its reading.
The topics where the brand is in the answer today and the ones the plan adds.
Where the brand has a page the models read and where it is absent.
The checks a cited page passes and a brand's own page does not yet.
Known for: what the brand should be named for
The models associate each brand with a few topics and they are not always the ones the brand wants. We read which topics you are named for against every rival and what the models associate you with and the gap between that and your positioning is the first layer. A brand that wants to be known for service and is named for price has a content problem before it has a page problem.
Present at: which pages at each step
People ask a different question at each step of their journey and the models draw on different pages to answer each one. The board shows for every topic and every step how much of what the models read is yours. The low cells are the briefs. A brand can own the Execute step with its switching page and be absent at Shortlist where the choice is made.
Ready to be read: what each page needs
A model reads a page for plain facts stated once and laid out the same way everywhere. We judge your pages and the pages cited instead of them for the same things. Structured facts from your fact sheet. Prices and terms stated. Proof in numbers and reviews. Headings in the words of the question. A date this year. We also check that your own pages agree with your fact sheet because a model that finds three versions of a price repeats one of them.
Owned and earned
Most of the pages a model reads about you are not yours. With search on a model draws on a handful of pages and repeats them and the list is usually a comparison site and a review site and a rival's page before your own. The plan covers those pages as well as yours. A page you own is fixed this week. A page you do not own gets a correction request with the fact sheet attached or a partner brief or a better page of your own for the models to find instead. Whether a fact is learned or found decides the fix and that reading is on the Data integrity page.
The deliverable
The plan is a list of briefs and not a deck. Each brief names the answer it fixes and the page it has to beat and the facts it carries.
Measured after every change
Every page behind the answers is compared with last week's copy. A change is logged and called trivial or material and the share of answers that recommend you is followed after it. That is how you know a page moved the answer and not the season or a new model version. The reading repeats on the schedule you choose and the plan is reordered from it each month.
How a change is told from chance is on the Data integrity page.
How a content strategy project works
The work follows a discovery reading or starts with one. We need your fact sheet and your positioning and the list of your own sites and the first plan lands within two weeks of the reading.
- We read the pages behind the answers. Every page the models cite for your category and every page of yours the finder holds is read and judged against the journey for its step and its point of view and its proof. The counting is done outside the AI.
- We write the plan. The three layers become a list of briefs ordered by the step where you are most missing. Each brief is drafted with AI from your fact sheet and your team or agency delivers it. Corrections and partner briefs go out for the pages you do not own.
- We measure after every change. Each change is logged and the recommendation share is followed after it and the plan is reordered each month from what moved. The routine is the team's and runs without us.
What we do differently
- We start from the pages the models read. Not from a keyword list. The pages the answers cite are read and judged against the journey and the gap between them and yours is the plan.
- The plan is a list of briefs. Each brief names the answer it fixes and the rival page it has to beat and the facts it carries. A strategy deck does none of that.
- Owned and earned together. Most cited pages are not yours. The plan covers corrections and partner and comparison pages as well as your own site.
- Measured after every change. Every page change is logged and called trivial or material and the share of answers that recommend you is followed after it.
Questions people ask us
What is an AI brand content strategy?
A plan for what a brand publishes and fixes so that AI models name it and describe it correctly. It says what the brand should be known for and which pages it needs at each step of the user's journey and what each page needs for a model to read it. It covers the pages the brand does not own as well as its own site and it is measured by whether the answers moved.
How is content for AI answers different from SEO content?
SEO content is written to rank one page for one query. A model samples every page it can find and repeats what it finds most often and most consistently. Content for AI has to state the same plain facts on every page and carry proof and fit any index and it has to exist at every step of the journey and not only where the traffic was.
Do we need to publish more content to be named by AI?
Usually not more. Usually the same facts stated consistently everywhere you already publish and a few pages at the steps where you are missing. Volume without consistency gives a model three versions of a price to choose from.
Can one good page change what ChatGPT says about a brand?
Rarely. A model does not read your best page once. It samples everything and repeats what it finds most often so one page is outweighed by the pages that disagree with it. The plan changes many pages a little and the earned pages too and that is what moves an answer.
Who writes the content in a Model Research content strategy?
Your team or your agency. We write the briefs and we can draft the structured data and the outlines with AI from your fact sheet. The people who know your pages and your channels deliver them.
How do we know a content change moved the AI answer?
Every page behind the answers is compared with the week before and a change is logged and called trivial or material. The share of answers that recommend you is followed after each change with its range so that a move inside the range is never called a result.
Our details are on the contact page if you would like the plan for one step of your journey.