How to Set Up Your Website for Google AI Max
22 May 2026
Conventional pages
Connected business objects
Surface pages stay. Underneath them, the business gets a connected knowledge graph that pages, ads, AI answers, and sales teams all draw from.
Your website is the second half of AI Max setup. The first half — conversion tracking, AI Brief, negatives, brand controls, bidding, experiments — is the agency side. This is the side that lives in the CMS, and it's the part most marketing teams under-think.
The sharper way to think about it: the AI-era website isn't a website and it isn't a CMS. It's a Business Knowledge Layer with a publishing system attached. Conventional pages stay. Underneath them, the business gets a connected graph of what it sells, who it serves, what problems it solves, what situations it appears in, and what proof supports it.
Companion piece to What Is Google AI Max? and How to Set Up Google AI Max for Maximum Results. Read those first. This one is the website chapter.
The CMS of the AI era isn't where you manage pages. It's where you manage the business knowledge that pages, ads, search engines, AI answers, and sales teams all depend on.
"Make the business machine-understandable"
First, skip the "feed Google training data" framing. It's loose, and it's not what's actually happening. What's actually happening is more specific — and more useful:
- Google crawls your pages
- Indexes the entities it can identify
- Retrieves them when a relevant query lands
- Classifies them against intent signals
- Matches them to the right ad surface, AI answer, or organic result
- Cites them in AI Mode, AI Overviews, and ChatGPT/Claude/Perplexity
Every one of those verbs is gated by how readable your business is. If your offerings, customer types, problems, and proof are scattered across inconsistent pages, every verb degrades. If they're consistently expressed as connected entities, every verb gets sharper.
The goal isn't to feed training data. It's to make the business machine-understandable.
Why this isn't a CMS rebuild
The reaction most marketing teams have when they hear "your website needs a knowledge layer" is panic — they assume rip-and-replace. Skip the panic. This is additive, not destructive.
Conventional pages stay. The homepage, About, service pages, pricing, case studies, contact — all of it still matters for human navigation. What changes is what those pages are made from. Instead of unique copy authored directly into each page, pages become compositions of underlying business objects.
Most modern CMSes — Webflow Collections, Sanity Schemas, Contentful Content Types, WordPress with the right setup, Shopify Metaobjects — already support this. You don't need a new platform. You need a new way to author content inside the one you have. The shift is conceptual before it's technical.
The MVP layer — five objects to start with
Don't try to model ten things on day one. Start with the five that pay back the fastest. These alone fix most of the inconsistency, improve service pages and landing pages, generate cleaner FAQ blocks, and give AI Max sharper signals to learn from.
Offering
What you sell — products or services as discrete entities. Each has a name, a description, the customer it's for, the problem it solves, the proof that supports it, and the price tier.
Customer Type
A distinct audience segment your offerings serve. Defined by industry, role, stage, geography, or size. Each Customer Type connects to specific Offerings, Problems, and Proof.
Problem
A specific problem your business is built to solve. Has severity, root cause, who it affects, and what happens if unsolved. The reason a Customer Type might engage your Offering.
Situation / Question
A specific scenario where your Offering applies — the user's actual context, often phrased as the question they'd ask. Atomic, reusable, schema-ready. The unit that AI Mode, ChatGPT, and Business Agent for Leads can answer with.
Proof
Evidence: a case study, a named result, a client testimonial, a third-party benchmark, an external citation. Tied to the Customer Type and Offering it validates.
Even the smallest version of this works. Three connected objects — one clear Offering, two or three Customer Types with what each cares about, one Proof per Customer Type — measurably improves AI Max text customisation. Because the system stops generating contradictory ad copy from inconsistent page sources.
The advanced layer — five more as the business matures
Once the MVP layer is producing — faster page builds, consistent copy, better AI Max signals — expand. The advanced layer is where expertise-led brands lock in differentiation.
Methodology
Your proprietary how-you-do-it. Reusable IP that distinguishes you from competitors selling the same nominal service. (For Kaliber: Single Purpose Campaigns, AI Max Readiness, Commercial Answer Pages.)
Comparison
How you stack against alternatives — vs a competitor, vs in-house, vs a different approach. Critical for evaluation-stage queries and AI answers that compare options.
Location
Geographic entities you serve. City, country, region, neighbourhood. Drives local schema, geo-targeted pages, and regional ad routing.
Author
The people behind the work — bylines, expertise statements, accountability. AI engines weight authorial credibility heavily when picking sources to cite.
Campaign / AI Brief
Your active AI Briefs and campaign messaging guidelines, modelled as content. Names the Offerings, Customer Types, and Proofs to surface for a specific commercial moment.
Methodology is first-class for expertise-led brands (consultancies, agencies, B2B services). For product-led brands, the equivalent object is Process (how it's made), Ingredients (what's in it), Technology (what powers it), or Product System (how products work together). Same slot, different name. The point is: your proprietary how-you-do-it deserves an object.
The operational layer — what makes objects real
Modeling the objects isn't the work. The work is what each object triggers. Without an operational layer, a knowledge layer is just a prettier database.
For every object you stand up, you should be able to answer:
What does it actually trigger?
An Offering that doesn't trigger pages, schema, internal links, FAQs, proof connections, AI Brief slots, an owner, and a measurement plan is a content artefact, not infrastructure. The discipline of answering all eight is what turns the knowledge layer into something that actually moves performance.
Why this is the website half of AI Max setup
Tying it back to the rest of the series:
- Text customisation generates ad copy from your pages. With a consistent knowledge layer feeding those pages, AI Max generates consistent ad copy. Without it, every page is a different draft and the AI inherits the inconsistency.
- Final URL expansion picks the page Google routes to. A clean knowledge layer means you can build precise pages for each Situation × Customer Type combination — and trust the routing instead of policing it.
- AI Brief becomes precise. "For the aesthetic-clinic customer type, prioritise the LP Testing offering and the [client] proof point" is a more useful brief than "promote our marketing services." That precision only exists if those objects exist.
- Business Agent for Leads answers user questions inside the ad, drawing on your site. Questions authored as atomic objects produce accurate, on-brand answers. Questions scattered across page copy produce contradictions.
The agency-side setup matters. The website-side setup matters too. AI Max amplifies both — and amplifies the gaps in both.
Common ways the knowledge layer goes wrong
- Treating it as SEO work, not business strategy work. SEO can use the output. It can't build the input.
- Modeling objects without changing the authoring workflow. Objects exist in the CMS. Nobody uses them. Pages still get hand-written. Same inconsistency, prettier schema.
- Skipping the relationships. Objects exist but don't reference each other — Offerings don't link to Customer Types, Problems don't link to Proofs. The graph stays flat. AI engines can't follow the connections.
- Inconsistent definitions across objects. Offering "Performance Partner" and a separate "Performance Marketing" service page describe the same thing two different ways. Models the conflict instead of resolving it.
- Going straight to all 10 objects, getting overwhelmed, doing none. Start with the MVP layer. Expand only when the first five are producing.
- No operational layer. Objects exist, no one knows what they power, schema doesn't update, FAQ blocks stay manual, AI Briefs don't reference them. A prettier database.
- Mistaking the CMS for the strategy. The CMS publishes the knowledge layer. It isn't the knowledge layer. Choose a CMS based on how well it manages, connects, governs, and publishes objects — not on which logo is trending.
The migration path
You don't need to model everything before you publish anything. Three phases, in order.
Audit (weeks 1-2)
- Inventory existing pages and the entities they implicitly reference
- Identify duplicated copy across pages
- Pick the MVP 5 objects to model first
- Identify the owner and approval flow per object type
Build (weeks 3-8)
- Set up the 5 object types in your CMS
- Author 10-20 instances of each
- Rebuild 2-3 high-value pages using object references instead of inline copy
- Map each object type to its schema and wire up auto-generation
Compose (ongoing)
- Add the advanced layer as the business matures
- Rebuild remaining pages as object compositions
- Make object-authoring the default workflow
- Run the operational 8 questions against every object quarterly
Frequently asked questions
Do I need to rebuild my CMS to do this?
No. Most modern CMSes — Webflow, Sanity, Contentful, WordPress with the right plugins, Shopify with metaobjects — already support collections, references, and reusable blocks. You don't need a new platform. You need a new way to author content INSIDE the platform you have. The shift is conceptual before it's technical.
Which objects should I model first?
Start with the MVP layer of five: Offering, Customer Type, Problem, Situation/Question, and Proof. These produce the biggest gain for the least taxonomy overhead. Once they're producing — faster page builds, cleaner copy, better AI Max signals — add the advanced layer (Methodology, Comparison, Location, Author, Campaign/AI Brief) as the business matures.
Is this just SEO content planning?
No. SEO content planning decides what pages to publish. The knowledge layer decides what business entities exist and how they relate. Pages are a downstream output. The same Offering object can produce a service page, a landing page variant, a blog post, a comparison page, a homepage hero, AI Brief input, and sales collateral — without anyone rewriting the core definition five times.
Does Methodology apply to every business?
It's first-class for expertise-led brands — consultancies, agencies, B2B services where how-you-do-it is part of the differentiation. For product-led brands, the equivalent object is Process (how it's made), Ingredients (what's in it), Technology (what powers it), or Product System (how products work together). Same slot, different name. The point is that your proprietary how-you-do-it deserves an object.
How does schema markup fit into this?
Schema is the expression layer, not a content object. Each object type in your knowledge layer maps to a specific schema type — Offering maps to Service or Product, Question maps to FAQPage, Proof maps to Review or Testimonial, Author maps to Person, Location maps to Place. When the knowledge layer is clean, schema generation becomes a downstream automation, not a manual chore per page.
Who in my team should own the knowledge layer?
Cross-functional, but with one accountable owner — usually the head of marketing or content strategy. The owner sets the object schemas and approval rules. Subject matter experts author Offerings, Methodologies, Proofs. Editorial authors Situations, Questions, Comparisons. Web teams build the CMS infrastructure. Without one accountable owner, the knowledge layer drifts and inconsistencies creep back in within a quarter.
What's the minimum viable knowledge layer to start with for AI Max?
Three objects connected: one clear Offering definition, two or three Customer Types with what each cares about, and a Proof point per Customer Type. Even that small graph improves AI Max text customisation noticeably — because the system stops generating contradictory ad copy from inconsistent page sources. From there, expand.