Every website hits a ceiling, and most owners try to break through it with the wrong tool. They pay for maintenance when they need growth — and pay for a rebuild when what they actually need is structure.
Boris Cherny, who leads Claude Code at Anthropic, recently posted a framework for how engineering, product, and design roles are evolving — five archetypes:
- Prototyper: comes up with brand new ideas; churns out many ideas, most of which don’t ship
- Builder: quickly turns a prototype/idea into production-grade product/infra
- Sweeper: cleans up the UI, simplifies the code and system, unships, optimizes performance
- Grower: takes a product that has been built and iterates on it to improve Product-Market Fit
- Maintainer: owns a mature system to make it secure, reliable, fast, and efficient as it scales
His point: people and teams cluster into two or three of these, regardless of job title — and a healthy team needs a different mix depending on where the product is.
I’ve been sorting client work into lifecycle stages for years — I used to call them Rescue, Build, Authority, and Maintain. Cherny’s five are cleaner and more precise, so I’m adopting them. Websites move through the exact same archetypes as products do.
The lifecycle, at a glance
Tap any stage to jump to what that work actually involves.
Most retainers only cover stage five. AI-era visibility is earned at stage four.
The five stages of a website
Prototyper. Concepting and iteration before commitment: wireframes, information architecture drafts, content models, the ideas that get tested and mostly thrown away. Skipping this stage is how sites end up launched around the wrong structure.
Builder. Turning the surviving prototype into a production-grade platform: the build, the core schema, the launch. This is where most web projects get all their budget and attention — and where most agencies stop showing up once launch happens.
Sweeper. Cleanup and simplification: Core Web Vitals, technical debt, bloat, whatever got broken by the last well-meaning person who touched it. Unglamorous. Invisible when done right — but skip it and everything downstream gets slower and harder to fix.
Grower. The stage almost nobody is budgeting for. Answer-first structure, entities, citations — the content work that turns "indexed" into "cited." This is where Answer Engine Optimization lives.
Maintainer. Monitoring, incident response, keeping performance scores from quietly rotting. Necessary, but it assumes the site is already doing what it needs to do — it just keeps it doing that.
The mismatch
Most agencies operate on a Maintainer retainer: uptime, speed, status quo. But if your site hasn’t been re-architected for the AI era, you don’t need a Maintainer — you need a Grower. You’re paying a mechanic to keep the car running while the real problem is the road moved.
AI didn’t penalize your site — it just stopped needing to click it to answer the question.
That mismatch is exactly why AI visibility looks like decline even when nothing technically broke. The site didn’t get worse. The bar moved — Google, ChatGPT, Perplexity, and the rest started answering questions directly instead of just linking out — and nobody rebuilt for the new stage.
A site that scores perfectly on a technical audit can still be functionally invisible to an AI answer engine, because a clean technical foundation (Sweeper-stage work) doesn’t automatically produce answer-first content structure (Grower-stage work). They’re different jobs, done at different times, and most retainers only cover one of them.
What a healthy mix looks like
Cherny made this point about product teams — the right mix of archetypes changes depending on the product’s stage. Same logic applies directly to websites:
- A new site needs Prototyper + Builder + Sweeper (even on greenfield projects, something always needs sweeping)
- A growing site needs Builder + Sweeper + Grower
- A mature site with strong visibility needs Sweeper + Grower + Maintainer
Most agencies are only staffed and scoped for Maintainer — regardless of which stage the site is actually in.
Figuring out which stage you’re actually in
Answer three questions and this will point you to the stage your site needs next.
For reference, those three checks in full:
- Technical health check. Run your site through PageSpeed Insights. If you’re consistently in the green for performance and Core Web Vitals, you’re probably past Sweeper.
- Answer-first check. Pull up your three most important pages. Do they answer the core question in the first two sentences, in plain readable text — not buried under a headline and three paragraphs of throat-clearing? If not, you haven’t done Grower work yet, regardless of how much content exists.
- Citation check. Ask ChatGPT, Perplexity, or Google’s AI Overviews something like “what’s the best [service] in [city]” or “how do I [problem]” — the kind of query that triggers an AI answer. If you don’t show up, that’s Grower-stage work waiting, not a Maintainer-stage bug to fix.
If you fail check 1, start there — nothing else matters until the foundation is stable. If you pass check 1 but fail 2 or 3, that’s the gap this post is about.
What this looks like in practice
I’ve worked with a long-standing local services client whose site was technically pristine — fast, clean, no errors — but had been sitting in pure Maintainer mode for years. Content hadn’t changed since launch. It ranked fine for branded searches and was invisible for everything else, including the AI-generated answers now showing up ahead of the search results.
The fix wasn’t a redesign. It was moving the site into Grower mode: restructuring key pages so the answer came first, adding the entity and schema signals that give AI systems something concrete to cite, and treating a handful of pages as living documents instead of finished ones. Within a few months, the site started showing up in AI-generated answers and citations it had never appeared in before — same technical foundation, same domain authority, just a different stage of work applied to it.
That’s the pattern worth watching for in your own site: technically fine, but frozen at whatever stage it was last touched.
If you’re unsure which stage your site is in, or you suspect you’re paying for the wrong kind of work, run the three checks above. If you pass the technical check but fail the citation check, you have a structural problem, not a technical one. Stop paying for uptime and start paying for relevance.
Frequently asked questions
What’s the difference between website maintenance and Answer Engine Optimization?
Maintenance keeps a site running the way it already runs: uptime, speed, security patches. Answer Engine Optimization (AEO) changes what the site says and how it’s structured, so AI systems like ChatGPT, Perplexity, and Google’s AI Overviews can cite it directly in an answer. A site can be perfectly maintained and still be invisible to AI search, because maintenance and AEO are different kinds of work.
Why did my site’s traffic drop even though nothing changed on it?
Usually because the site didn’t change — but search did. AI answer engines increasingly answer questions directly instead of sending a click to a website. If a site was built for the traditional “rank and get clicked” model but never restructured for answer-first content, it can lose visibility even with a stable technical foundation and no errors.
Is schema markup (like FAQPage JSON-LD) enough to get cited by AI?
No. Schema functions as entity infrastructure — it helps AI systems understand what a page is about — but it doesn’t independently drive citations. Visible, answer-first content structure is the active ingredient; schema supports it rather than replacing it.
How do I know if my website needs a rebuild or just Grower-stage work?
Run a technical health check first. If Core Web Vitals and performance are already solid, you don’t need a rebuild — you need Grower-stage work: restructuring key pages so they answer the core question up front, and adding the entity signals that make the content citable.
How often should a website move through these five stages?
There’s no fixed schedule. Prototyper and Builder happen once per platform. Sweeper is needed at the start, after neglect, and periodically as debt accumulates. Grower is ongoing as content and AI search evolve. Maintainer runs continuously in the background once the other stages are in good shape.