The AI Readiness Maturity Model

Search for an AI search maturity model and you'll find several. Conductor has a five-stage matrix running from Reactive to Agentic. Brave Bison has four stages, from Reactive to Predictive. Superlines has four levels, from Ad Hoc to AI-First. They're thoughtful, and they share one thing: they measure your organisation. Who owns AI search, how decisions get made, whether it shows up in reporting. Conductor says it directly: its matrix looks at "how teams are structured, how decisions get made" (Conductor).

That's a real question, but it isn't the first one. A company can sit at a mature "strategic" stage, with an owner, a budget and a weekly report, while running a website that AI crawlers aren't allowed to open. Every stage of the team model can be in place while the website itself never got past the starting line.

This post is the other half: a maturity model for the website. Four stages, each with a test you can run yourself, a clear sign you've passed it, and a list of things that are a waste of money until you have. It's written for a small or mid-sized business working out where its site actually stands, not for a team planning an org chart.

Why a Website Needs Its Own Model

A team maturity model works on a scale. You're a bit more mature in one area, a bit less in another, and the overall picture is a blend. A website doesn't work like that with AI, because the steps happen in order and each one depends on the last.

An AI system can't understand who runs a page it was never allowed to read. It can't cite a business it can't tell apart from three others with similar names. Each stage is a gate: until it's open, nothing beyond it counts for much, however well that later work was done.

That has one practical consequence that's easy to miss. Your stage is set by the lowest gate you haven't passed, not by your average. A site can have excellent business details, a clean set of structured data and well-written answers to customer questions, and still be at stage zero because a firewall rule blocks the crawlers that would read all of it.

This is also how to read an AI readiness score next to a stage. Our own score adds up four parts, and the first one, whether crawlers can reach your site, is worth 15 of the 100 points (our post on what moves the score breaks down the split). If your site can't be reached at all, everything after that fails too and the whole score falls with it. But the more common case is narrower: the site loads normally, and only AI crawlers are turned away by a robots.txt rule or a bot filter. Then the rest of the site can still be checked and scored, and the total can look respectable while the gate that matters most is shut. The score tells you how much work is left in total. The stage tells you which piece of that work is blocking everything else. You need both, and when they seem to disagree, the stage tells you what to do first.

The Four Stages at a Glance

Stage	What it means	The gate to the next stage	Mostly in your control?
0. Invisible	AI systems can't reach or read your key pages	Crawlers can reach the pages and get real content from them	Yes
1. Found	AI can read your pages, but can't reliably tell who you are or what you do	Your site states clearly and consistently who you are, and answers the basic questions about you	Yes
2. Understood	AI can describe your business correctly from your own site	AI answers start citing or naming you, steadily, over time	Partly
3. Cited	AI answers use you as a source or recommendation, at a rate you can measure	No gate; this stage has to be kept, not passed	Partly

The first two gates are entirely about your own site, which makes them the cheapest and most certain work you'll do. The third depends partly on things you don't control: other websites, the AI systems themselves, and plain chance. That's why the order matters so much. The work you control comes first, because the stage you only partly control is built on top of it.

Stage 0: Invisible

What it looks like: AI assistants either don't know you exist or describe you from old, secondhand information, because they can't get to your current pages. From the inside, everything looks fine. The site loads in your browser, Google may even rank it, and nothing tells you that AI crawlers are being turned away or handed an empty page.

The usual causes: a robots.txt rule that blocks AI crawlers, sometimes added years ago or switched on by a hosting or security setting. A firewall or bot-protection service that blocks automated visitors without anything showing up in robots.txt. Pages marked noindex by mistake. Or a site built so that the content only appears after scripts run in a browser, which many AI crawlers don't do, so they receive a near-empty page.

How to tell you're here: check four things on your homepage and your two or three most important pages.

1. Does robots.txt let AI search crawlers in? Open yoursite.com/robots.txt and look for rules naming crawlers such as OAI-SearchBot, PerplexityBot or Claude-SearchBot, or a blanket rule that blocks every crawler.
2. Is the page allowed into search at all? Look for a noindex instruction in the page's code or headers. For Google's AI features specifically, this is the whole entry ticket: Google says a page "must be indexed and eligible to be shown in Google Search with a snippet," and that "there are no additional technical requirements" (Google).
3. Is the content there without scripts? View the page source (not the inspected page) and search for a sentence from your main content. If it isn't in the source, a crawler that doesn't run scripts won't see it either.
4. Is anything in between blocking bots? If you use a firewall or bot-protection service, check whether it challenges or blocks known AI crawlers.

Our post on whether AI crawlers can reach your site goes through each of these in detail.

What moves you up: fixing whichever of those four failed. This is usually the quickest stage to leave. Most of the causes are a single setting or a single line, and once fixed, you're no longer invisible the next time a crawler visits.

What's a waste of money here: almost everything else. A visibility tracking tool will just report zeros, week after week, and tell you nothing about why. New content written "for AI" won't be read. Structured data won't be seen. Getting listed in more directories helps a little, because AI can learn about you from those, but it's working around a locked door instead of opening it.

Stage 1: Found

What it looks like: AI crawlers can reach and read your pages, but the answers AI gives about you are vague, mixed up, or wrong in small ways. It might get your service area wrong, mix you up with a business with a similar name, describe last year's services, or simply not know what you do beyond your name. If you ask an assistant about your business by name, the answer sounds like it was assembled from fragments, because it was.

Why it happens: reading a page is not the same as knowing who's behind it. Most sites never state plainly, in one consistent place, who the business is: its exact name, what it does, where it works, how to reach it, and which outside profiles belong to it. The facts are scattered across a footer, an About page and a contact form, sometimes with different versions on different pages, and the AI has to guess which ones go together.

How to tell you're here: two tests.

The identity test. Can an AI system find one clear, consistent statement of who you are? That means your business name, what you do, your location or service area, and your contact details, stated the same way across your site, ideally in structured data as well as in visible text, and matching your profiles elsewhere. Check for the boring problems first: an old phone number on one page, a different trading name in the footer, an address that changed but not everywhere.

The closed-book test. Using only your own website, could a stranger answer six basic questions: what the business does, where it's based, which areas it serves, what services or products it offers, who runs it, and how to contact it? If any answer requires knowing something that isn't written down on your site, an AI working from your site can't answer it either. Our post on whether AI can answer questions about you explains how to run this properly.

What moves you up: making the facts about your business explicit, consistent and connected. In practice that means one clear business identity, stated in both visible text and structured data, with the same details everywhere, connected to the pages it runs and to your outside profiles. Our post on whether AI can tell who runs your site covers how that connection works and the ways it quietly breaks. Then fill the gaps the closed-book test found, usually a missing service area, a vague description of what you actually do, or no named person behind the business.

What's a waste of money here: buying a visibility tracker with dozens of questions. It will tell you that you're rarely mentioned, which you can already guess, without telling you the reason, which is on your own site. Rewriting your whole site in an "AI-friendly" style also comes too early, because clearer prose doesn't help much if the AI still can't tell whose prose it is. And piling on extra types of structured data that don't describe your business adds weight without adding meaning.

Stage 2: Understood

What it looks like: when you ask an AI assistant about your business by name, the answer is accurate. It knows what you do, where you work and how to reach you, and when it searches, it can find your own pages to back that up. But when someone asks a general question, "best [service] in [city]", "who can help with [problem]", you're named only sometimes, or not at all.

Why that's normal: being understood makes you eligible to be recommended. It doesn't make you recommended. At this stage, whether an AI names you depends on how your site compares with every other site that answers the same question, and on what the rest of the web says about you: reviews, directories, articles, lists. Those are partly outside your control.

How to tell you're here: you pass both Stage 1 tests, and when an AI assistant searches for your business by name, it describes you correctly and shows your own pages among its sources. Run that check in a temporary chat with personalisation off and web search switched on, so the answer isn't shaped by your own account history or location. Asking by name is a generous test, because the question already hands the AI your name. That's fine here, since Stage 2 only claims that AI can find and describe you. It says nothing yet about whether AI would recommend you to someone who hasn't heard of you, which is what Stage 3 measures.

What moves you up: two things, and only one of them is about your own site.

• On your site: pages that answer the specific questions customers actually ask, clearly and in plain terms, with the details that make you the right choice for a particular situation. A general "our services" page competes poorly with a page that answers the exact question someone typed.
• Off your site: showing up in the places AI systems read when they answer questions in your category. That means accurate listings in the directories that matter for your trade, reviews on the platforms your customers use, and being included in the lists and articles that come up as sources. When you check an AI answer that doesn't name you, the sources it does show are a ready-made list of where to look.

What's a waste of money here: treating any single AI answer as a verdict. Answers vary from one ask to the next, so a single "it didn't mention us" proves nothing. This is also the stage where people start chasing a "ranking position" in AI answers. That order changes too much from one ask to the next to be worth paying anyone to move.

Stage 3: Cited

What it looks like: AI assistants name you, or list your site as a source, for a meaningful share of the questions that matter to your business. Not every time, because no business is named every time, but steadily enough that it shows up as a rate rather than the occasional lucky answer.

Why it isn't a finish line: this is the one stage you can fall out of without doing anything wrong. A competitor publishes a better page. A directory drops your listing. The AI system changes how it searches, or which sources it prefers. Your rate can drop for reasons that have nothing to do with your own site.

How to tell you're here: you need a measurement, not a feeling. Pick a fixed set of questions your customers would ask, ask each one several times, and track how often you're named or cited, round after round. Our guide on how to track brand mentions in ChatGPT covers how to run that routine and how to tell a real change from normal variation.

What keeps you here: two habits. Watch your own site for anything that silently knocks you back down a stage, which is covered in the next section. And keep measuring, so that when your rate drops, you see it early and can check whether the cause is on your site, on someone else's, or inside the AI system itself.

What's a waste of money here: reacting to every wobble. Once you're cited, most week-to-week movement is noise. A drop that's spread across many questions and lasts more than one round is worth acting on. A single bad week isn't.

Sites Slide Back Down

The stages aren't permanent, and the most common way to lose one has nothing to do with AI at all. It's ordinary website maintenance.

A redesign ships with a template that leaves out your business details. A plugin update switches off your structured data. A new security setting starts blocking crawlers it doesn't recognise. A developer adds noindex to a staging site, and it comes along to the live site at launch. A page that held your service area gets merged into another and the details are lost along the way.

Any of these can drop a Stage 3 site straight back to Stage 0 or Stage 1 overnight, and nothing about how the site looks to a visitor will tell you. This is the strongest reason to keep checking the gates you've already passed, not just the stage you're working on. Our post on the AI readiness loop describes the check, fix and monitor cycle that catches this, so a slide back down is caught within days rather than months.

Find Your Stage in Five Minutes

Answer these in order, and stop at the first "no". Your stage is the one where you stopped.

Stage 0 gate

1. Does your robots.txt allow AI search crawlers onto your important pages?
2. Are your important pages free of noindex?
3. Does your main content appear in the page source, without needing scripts to run?
4. Is no firewall or bot-protection service blocking AI crawlers?

If any answer is no, you're at Stage 0: Invisible. Fix that first.

Stage 1 gate5. Is your business name, what you do, where you work and how to reach you stated the same way across your site? 6. Is that identity also stated in structured data, and does it match your outside profiles? 7. Could a stranger answer six basic questions about your business using only your site?

If any answer is no, you're at Stage 1: Found. Make your business facts clear and consistent.

Stage 2 gate8. When you ask an AI assistant about your business by name, in a clean check, is the answer accurate? 9. Across a fixed set of customer questions, asked several times over a few weeks, are you named or cited at a rate that holds steady?

If question 8 is a yes and question 9 is a no, you're at Stage 2: Understood. Work on specific answer pages and your presence on the sites AI reads.

If both are yes, you're at Stage 3: Cited. Keep measuring, and keep checking the earlier gates after every site change.

Using This Alongside a Team Maturity Model

The two kinds of model answer different questions, and they're more useful together than either one alone.

A team model tells you whether your business is organised to do the work: whether someone owns it, whether it's measured, whether it has a budget. A website model tells you which work to do next. Put them side by side and four situations appear:

• Mature team, early-stage site. The most expensive combination. There's budget, reporting and tools, often including paid visibility tracking, all measuring a site that hasn't passed the first or second gate. The fix is not more process. It's pointing the existing process at the gates.
• Early-stage team, mature site. Common in small businesses that got the basics right without ever thinking about "AI search" as a programme. This is fine. The site is doing the work. The risk is only that nobody's watching for it to slide back.
• Early-stage team, early-stage site. Start with the site. The first two gates are cheap, quick and need no team structure at all, just someone willing to fix a handful of settings and pages.
• Mature team, mature site. The team model is now the one that matters most, because what's left is mostly Stage 3 work: off-site presence, measurement, and reacting well to change.

Conductor makes a related point from the other direction. Writing about a team's SEO foundation, it says "that foundation typically only gets you to stage two" of its team model. The website model is a closer look at that foundation, and at why it has to come first.

Common Mistakes

Averaging instead of gating. Treating your position as a blend of strengths and weaknesses hides the one gate that's actually blocking you. A site with excellent business details and blocked crawlers is still invisible.

Buying measurement before passing the first two gates. A visibility tool can tell you you're not being cited. It can't tell you that your robots.txt is turning crawlers away or that your business details contradict each other. Those are free to check and they come first.

Assuming a good-looking site has passed Stage 0. The causes of invisibility rarely show up to a human visitor. A beautiful site that renders everything with scripts, or sits behind an aggressive bot filter, can be invisible to AI while looking perfect in your browser.

Treating Stage 3 as permanent. Being cited is a rate that has to be kept, not a level you reach once. It can drop because of a competitor, a source, or a change in the AI system, with nothing wrong on your side.

Never rechecking the gates you've passed. The biggest risk to a mature site is ordinary maintenance. A redesign, a plugin update or a new security setting can send you back to the start without anyone noticing.

Skipping straight to Stage 2 work. Writing pages that answer customer questions is valuable, but only once AI can read your site and tell whose answers they are. Done in the wrong order, it's good work that nobody sees.

FAQ

What is an AI readiness maturity model?

It's a way of describing how far along a website is in being usable by AI systems, in stages that have to be passed in order: first AI can reach and read the site, then it can tell who the business is and describe it correctly, then it starts naming or citing it in answers. Most published AI search maturity models instead describe how mature a company's team and processes are.

How is this different from an AI readiness score?

A score adds up how well the site does across several areas, so strong results in one area can balance weak results in another. A stage is set by the first gate the site hasn't passed. Both are useful: the score shows how much work is left, the stage shows what to fix first.

What stage are most websites at?

We haven't published figures on that, so we won't guess. In practice, many sites that have never been checked turn out to have at least one Stage 0 or Stage 1 problem, often an old crawler rule or business details that don't match from page to page. The five-minute check above will tell you where yours is.

Can different pages be at different stages?

Yes, and it's common. Your homepage might pass every gate while your service pages are built in a way crawlers can't read, or a whole section was marked noindex during a rebuild. Run the five-minute check on the pages that matter most for getting customers, not only the homepage. Your site's real stage is the stage of those pages.

Is Stage 3 the same as ranking first in AI answers?

No. Being cited means you're named or used as a source in a steady share of answers. Where you appear inside any one answer changes too much from one ask to the next to treat as a position, so Stage 3 is measured as how often you show up, not where.

Do I need structured data to get past Stage 1?

It's the clearest way to state your business identity in a form machines read reliably, which is why the identity test includes it. It isn't a special requirement for Google's AI features, which Google says need no particular structured data. The point of Stage 1 is that your business facts are clear and consistent everywhere, and structured data is one of the strongest ways to do that.

Does a mature team model make this unnecessary?

No. A team can be well organised and still be working on a site that's stuck at Stage 0 or 1. The team model tells you whether you can do the work. The website model tells you which work comes first.

Want to know which gate your site is stuck at? A free AI readiness check looks at crawler access, page readability and your business identity in one pass, and lists the specific problems standing in the way.