Posted by
Sam Beddall
This blog explores the findings and insights from a recent webinar, which you can watch or listen to here: Is your website invisible to AI search?.
If you are planning a new website, you are probably already asking the familiar questions.
Will it look good? Will it be easy to update? Will it convert? Will it support SEO? Will it make us look like the serious, useful and differentiated business we believe ourselves to be?
All good questions. But they are no longer enough.
Your website now has another audience. Human buyers still matter, obviously. But AI systems, search crawlers and answer engines are also reading, interpreting and recommending businesses. Increasingly, B2B buyers are using tools like ChatGPT, Claude and Perplexity to research suppliers, compare vendors and build internal business cases before they ever speak to sales.
We are already seeing this at Articulate. Prospects have come to us after asking Claude or ChatGPT to recommend marketing agencies with specific expertise. That is interesting, useful and slightly unnerving, especially when they arrive with a credible-looking AI-generated diagnosis of their own marketing problem.
So the question for your next website is not only whether people can understand it. It is whether AI systems can access it, parse it, understand it, trust it and recommend it.
That means you need better questions for your website developer.
What does AI-ready actually mean?
Every website proposal will soon claim the new site is AI-ready.
The phrase sounds reassuring, but it needs interrogating. AI readiness is not a single feature, plug-in or line item. It is not something you can safely tick off because someone added an LLMs.txt file or wrote ‘AI search optimisation’ in the proposal.
A genuinely AI-ready website needs to work across three layers.
First, AI systems must be able to access and read the site. If they cannot reach your content, the rest is academic.
Second, the information needs to make sense to a machine. Clear copy, structured data, well-formed HTML and properly maintained source-of-truth pages all matter here.
Third, you need some way to measure whether any of this is working. That means looking at crawler behaviour, AI referrals, visibility, citations and the quality of AI-influenced traffic.
There is also a fourth, more human layer. You need to know whether the people building the site understand how quickly this field is changing, how they are using AI in development and what they will advise you not to buy.
So, before you sign off your next website proposal, ask these 10 questions.
1. Can AI systems actually access the site?
This is the first and most basic question.
If AI crawlers cannot reach your website, they cannot read, understand or cite it. The failure is often silent. A setting in robots.txt, a security rule, a badly configured firewall or an overzealous crawler-blocking tool can stop AI systems from accessing important content without anyone noticing.
That changes whether your company can appear in AI-assisted research at all.
Ask your developer how they will test crawler access before and after launch. Ask which bots they are allowing, which they are blocking and why. Ask whether the site’s robots.txt file has been reviewed specifically for AI crawlers, not just conventional search engines.
You do not necessarily have to allow every bot everywhere. Some organisations may have good reasons to restrict certain tools. But it should be a deliberate decision, not an accident.
The wrong answer is, ‘Don’t worry, we do SEO.’
The better answer is, ‘Here is how we check whether AI crawlers and search crawlers can access the pages you want them to read, and here is how we monitor that over time.’
2. Is our content being served in a machine-readable way?
A human sees a finished webpage.
A machine sees pieces arriving. HTML, text, fonts, images, JavaScript, CSS and other components all need to be assembled into something meaningful.
Matthew compares it to an IKEA kit. The contents arrive in a box, but somebody still has to build the sideboard. Some website technologies make that assembly easy for machines. Others make it much harder.
Many modern websites rely heavily on JavaScript, animation, dynamic rendering, embedded content and complex front-end frameworks. Those things can create beautiful experiences for human visitors, but they can also make important content harder for AI systems to access.
Google can handle JavaScript better than many other systems, but that does not mean every AI crawler can. As of the webinar, Matthew’s warning was blunt: major AI bots such as GPTBot, ClaudeBot, PerplexityBot, Meta and ByteSpider should not be expected to wait patiently for your JavaScript-heavy page to load and render.
So ask your developer how the site will serve important content.
Will core page content be available in clean, standards-compliant HTML? Will important copy be hidden inside scripts, animations, images or embedded video? Will AI systems be able to read the page without behaving like a fully interactive browser?
A site that looks impressive but hides the message from machines is not AI-ready.
3. Will the site be fast enough for AI crawlers?
Page speed has mattered for years because it affects user experience, conversion and search visibility. AI makes it even more important.
Slow pages frustrate human visitors, but at least you can often see that problem in analytics, user testing or Google Search Console. AI tools behave differently. If a request times out, they may simply leave. You may not know they failed to access the page.
There may be no obvious warning, no neat report and no ‘Claude Search Console’ telling you that your website was too slow to be considered.
This is why performance cannot be treated as a nice-to-have after launch. It needs to be part of the design and build approach from the beginning.
Ask your developer what page weight, load time and performance standards they are aiming for. Ask how they will test those standards, and ask what trade-offs they will make between visual design, tracking scripts, animation, images and speed.
At Articulate, we are actively looking at ways to reduce page weight, including replacing bitmap images with vectorised alternatives where appropriate. A well-compressed image might be 200 or 300 kilobytes. A vector version might be 10 or 20 kilobytes. On one site, we reduced average page weight to around 300 kilobytes by stripping out unnecessary code and keeping the build lean.
That is not only good for AI but also good for people.
Fast, accessible, standards-compliant websites are not a niche technical obsession. They make the experience better for everyone.
4. Will AI understand who we are and what we do?
Once AI systems can reach and read your site, the next question is whether they can understand it.
This is where copy and positioning become part of AI readiness.
If your homepage is full of vague claims, invented category language and inflated adjectives, you are making the machine work harder. You are also making life harder for human buyers, which is rarely a good sign.
Your website needs to say clearly what you do, who you do it for, what problems you solve and why you are different. That information should be easy to find on the homepage, about page, service pages and other core pages.
This does not mean turning the site into a dull technical manual. It means making your source material explicit enough for both people and machines.
You can still have a strong brand voice. You can still be interesting. You can still use narrative. But somewhere on the site, preferably in several obvious places, the truth of the business needs to be stated plainly.
What do you sell? Who buys it? What category are you in? Which sectors do you serve? What proof do you have? What should AI systems and buyers understand about you?
If the answer is buried under ‘transformational solutions for future-ready organisations’, you may have a problem.
5. How will schema be built into the site?
Schema is one of the most practical ways to help machines understand your website.
Think of it like barcoding a warehouse. Without labels, someone might still find the right item eventually, but it takes longer and creates more room for error. With structured labels, machines can identify what information means more quickly and confidently.
Schema can help explain things like:
- Your organisation
- Products and services
- Articles
- Authors
- FAQs
- Events
- Reviews
- Videos
- Locations
- Contact details
Adding schema as an afterthought is better than doing nothing, but the stronger approach is to weave it into the design and build from the start.
Ask your developer which schema types they recommend and how they will be implemented. Ask whether schema will be consistent across the site, not just added to a handful of pages. Ask how it will be maintained when content changes.
This is where AI readiness and good website craft overlap.
Schema is not a magic answer. But if you want machines to understand your content, it is a sensible way to serve meaning on a plate rather than hoping they infer everything correctly.
6. Where will our source of truth live?
AI systems can cite and summarise information from places you do not control.
That might include Reddit, review sites, directories, YouTube, old press coverage, social profiles, partner pages and articles on other websites. Some of that may be useful. Some of it may be inaccurate, outdated or incomplete.
You cannot control the whole internet, irritatingly. But you can make your own website a better source of truth.
This is where many organisations fall down. Their website no longer reflects the business they are actually in. Services have changed. Pricing has changed. Sectors have shifted. The positioning has moved on. Case studies are old. Press pages are unhelpful. Investor information is hard to navigate. Product details live in scattered PDFs. That creates risk.
If your truth is fragmented, AI systems may pick up the wrong version. They may cite an old price, an outdated description or a third-party summary that no longer reflects what you do.
Ask your developer and internal team where the canonical information will live.
Which pages should AI systems trust for company information? Where will current service descriptions live? How will old content be reviewed? Who owns the accuracy of pricing, product detail, sector claims and proof points?
This is a governance problem rather than just a technical one.
A new website can launch with accurate information, then drift slowly out of date. The more pages you have, the more content debt you create. Page sprawl is not just messy for users. It is messy for machines too.
7. How will we measure AI visibility?
AI visibility is still an emerging field, so beware anyone who claims to have a perfect answer.
There are now tools that run prompts against AI systems and report whether your brand appears. These can be useful, but they are also sensitive to prompt wording, persona, region, model choice and small changes in phrasing.
A visibility score might be directionally interesting. It is not always definitive.
That does not mean you should ignore measurement. It means you should treat it as a developing practice and build a sensible baseline.
Ask your developer or marketing partner what they will measure before launch and after launch.
That might include:
- AI referral traffic
- Chatbot referral sources
- Crawlers hitting the site
- Server logs where available
- Indexed pages
- Branded AI mentions
- Unbranded AI visibility
- Assisted conversions
- Lead quality from AI-influenced traffic
- Changes in core search visibility
- Performance data
- Key page accessibility and crawlability
HubSpot and other platforms are beginning to include AI visibility and referral reporting, and this will probably become part of normal SEO and website reporting over time. If you want a more joined-up view of measurement, automation and optimisation, see our work on HubSpot automation and AI.
For now, the important thing is to get a baseline before launch. Otherwise, you will not know whether the new website improved anything.
8. How will we know if AI recommends us?
Some AI-driven visits are easy to see. A person clicks a link from ChatGPT, Claude, Perplexity or another tool, and your analytics platform records the referral.
But that is only part of the story.
Someone may ask an AI tool for recommendations, see your name, then search for you separately. They might type your URL directly. They might ask a colleague. They might return later from a different device. That activity will not always be neatly tagged as an AI referral.
This means AI recommendations can create a small visible signal and a larger invisible influence.
Ask how your team will interpret that.
If you see 50 or 100 visits a month from chatbot referrals, that may be the measurable tip of a larger behaviour change. Those visitors may also be better qualified because they have already done some research before arriving.
This is one reason AI traffic can be valuable even if the raw numbers look small compared with organic search.
The question is not only how many visits came from AI. It is whether those visitors behave differently. Do they convert at a higher rate? Do they arrive further into the buying journey? Do they ask better questions? Do they move faster into qualified conversations?
Measure the volume, but pay attention to the quality.
9. How are you using AI to build the site, and where are the controls?
Any serious website developer should now have an answer to this question.
AI can be useful in software development, content operations, testing, code review, documentation, prototyping and process automation. Used well, it can help teams build better, faster and more reliable websites.
Used badly, it can create security, maintainability and accessibility problems.
The issue is not whether your developer is using AI. The issue is how.
You want to hear a grown-up answer.
How is AI used in the development process? Which tasks are AI-assisted? Who reviews the output? How is code tested? How are security risks checked? How do they make sure AI-generated work is maintainable, integrated with the rest of the site and aligned with accessibility standards?
This matters because vibe-coded sections and AI-generated code can look fine on the surface while hiding problems underneath. AI-assisted code can contain vulnerabilities, and some AI-built apps have shipped with serious security weaknesses.
Ask for the checkpoints. Ask about human oversight. Ask about testing. Ask who is accountable. Ask how issues are found and fixed.
AI should make the work stronger, not just faster.
10. What has changed in your approach recently, and what will you tell us not to buy?
AI is moving too quickly for last year’s answer to be good enough.
Ask your developer what they have changed in the last six months. Not what they have read. Not what they plan to explore. What they have actually changed in their process, tooling, testing, standards or recommendations.
If they answer that nothing much has changed, that tells you something.
A credible partner should be experimenting, learning and adapting. They should have a view on what has become important, what has become less important and what is not yet worth serious investment.
They should also be willing to tell you what not to buy.
This is a useful test because AI anxiety creates spending opportunities. It is easy to buy tools, plug-ins and monitoring products because they make you feel as though the problem is being solved.
Some may be useful. Some may be premature. Some may be expensive distractions.
Matthew’s current view is that heavy investment in LLMs.txt files, real-time translation of HTML into markdown and some AI citation monitoring tools may not be the best place to start. They may have some utility, but they are not a substitute for the basics.
A better investment is a fast, standards-compliant, well-written website with clear positioning, strong copy, structured schema and maintained source-of-truth content. In other words, the foundations of a good B2B website, not a bag of shiny AI tricks.
That is less shiny than buying a new AI visibility tool.
It is also more likely to help.
Changing your mindset
The important thing is not to treat AI readiness as a box to tick.
It is a change in how we think about the website. Your site is no longer only a destination for human visitors. It is also source material for AI systems that may influence what buyers see, believe and shortlist.
That does not mean chasing every new AI tactic or buying every tool with a clever dashboard. It means doing the fundamentals properly, with the AI audience in mind.
Make the site accessible. Make it fast. Make it parseable. Make the copy clear. Add schema. Maintain your source of truth. Measure what you can. Watch how the field changes. Ask better questions of the people building it. That is how you build a proper Difference Engine® rather than just a prettier website.
Most competitor websites will not do all of this well.
Some will still be stuck in a pre-AI model of website design. Some will look impressive but serve important content in ways machines struggle to read. Some will make vague claims and hope AI systems work out the rest.
That creates an opportunity for you.
The 10-question checklist
Before you approve your next website proposal, ask:
- Can AI systems access the site?
- Is our content being served in a machine-readable way?
- Will the site be fast enough for AI crawlers?
- Will AI understand who we are and what we do?
- How will schema be built into the site?
- Where will our source of truth live?
- How will we measure AI visibility?
- How will we know if AI recommends us?
- How are you using AI to build the site, and where are the controls?
- What has changed in your approach recently, and what will you tell us not to buy?
The answers will tell you a lot.
If your developer can answer clearly, specifically and with evidence, that is a good sign. If the answers are vague, adjective-heavy or overly certain in a field that is still evolving, keep asking. For more on making your expertise clear, useful and recognisably yours, read our guide to defining your tone of voice, or explore more insight on the Articulate blog.
If you’re having second thoughts about your developer or just want a friendly website chat with a multi-decade expert on the subject, book a 30-minute no BS call with our CEO Matthew Stibbe.
