Why is my website not getting mentioned in AI conversations?

Your website was built for a search landscape that will not exist in the near future.

To get mentioned in AI conversations and recommendations, you need to ensure that absolutely everything on your website is built and optimised correctly. Ambiguity, inference and missed messaging in your content can also stop you getting cited in AI.

Get help with this

The time to do something about this is now – not tomorrow. You know what can happen if you are late moving on this – you miss the boat, and this can have a significant impact on your business.

• What AI recommendations actually are.
• Your site is not being crawled properly.
• Your content does not answer real questions.
• You have no authority signals.
• Your HTML is a mess.
• You are not mentioned anywhere else.
• Your content is too thin or too vague.
• You have no schema markup.
• Your site is too slow.
• What to do about it.

A client asked me recently why their business never comes up when their prospects use ChatGPT or Perplexity to find suppliers in their sector. They had a website. They had an agency looking after their SEO. They had been in business for over a decade. And yet, when their target audience asked an AI assistant to recommend firms in their field, they were completely invisible.

It is a question I am getting more and more often, and the answer is rarely one thing. It is almost always a combination of structural problems that, individually, would not sink a site – but together, they are enough to keep you off the radar entirely when AI models are deciding who to mention.

What AI recommendations actually are.

Before we get into the reasons, it helps to understand what is actually happening when an AI assistant recommends a business.

Tools like ChatGPT, Perplexity, Claude, and Google’s AI Overviews do not have a database of recommended businesses they browse through.

They synthesise information gathered from web crawling, training data, third-party citations, and – in the case of tools with live search – current web results.

When someone asks “who are the best HR consultancies in the Midlands?” the AI is not running a directory lookup.

It is drawing on everything it has absorbed about those businesses from across the web and deciding which ones it has enough consistent, credible information about to confidently name.

That is an important distinction. Being recommended by AI is not about gaming an algorithm. It is about being genuinely well-documented, well-cited, and clearly understood by machines that read your site and everything written about you.

Your site is not being crawled properly.

AI crawlers – Anthropic’s ClaudeBot, OpenAI’s GPTBot, Google’s various crawlers – visit your website and try to understand what you do, who you serve, and why you are credible. If they cannot crawl your site properly, none of the rest matters.

Common crawl blockers include a poorly configured robots.txt file that inadvertently blocks AI crawlers, a slow server that times out before the crawler finishes, JavaScript-rendered content that crawlers cannot read, and pagination or filtering that creates content the crawler never reaches.

Some sites also actively block AI crawlers – which is a choice some businesses make deliberately – but if you are doing it accidentally through an overzealous robots.txt, you have already lost before the conversation begins.

Check your robots.txt file. Check whether GPTBot and ClaudeBot are explicitly blocked. Check your crawl stats in Google Search Console to see whether bots are hitting errors or being turned away.

Your content does not answer real questions.

This is the big one for most B2B sites.

A lot of B2B websites are written for the business, not for the buyer. They are full of statements about how experienced, passionate, and client-focused the company is, and entirely empty of anything that helps a prospect make a decision or understand a subject.

AI models are trained to surface content that answers questions. If your site does not contain answers – to real questions your buyers are asking – there is nothing there to pull from. You will not get cited because there is nothing citable.

Think about what your prospects are actually searching for. Not “HR consultancy Warwickshire” as a keyword, but the questions behind that search: what does an HR consultant actually do for a small business? What does it cost? When do I need one rather than doing it myself? How do I choose between them?

If your website answers those questions clearly and in plain English, AI tools have material to work with. If your website just says “we provide expert HR support to growing businesses”, they have nothing.

You have no authority signals.

Google has been talking about E-E-A-T for years – experience, expertise, authoritativeness, and trustworthiness. AI models care about the same things, because they are trained partly on the same sources Google values.

Authority signals for a B2B site include named authors with visible credentials, case studies with real outcomes, client testimonials that are specific rather than generic, references from credible third-party sources, press coverage, industry body memberships, awards, and published thought leadership that demonstrates genuine expertise.

If your website has no named people on it, no case studies, and no external validation of any kind, you look like a generic supplier. AI tools will not recommend a generic supplier when there are clearly-documented alternatives in the same category.

The fix here is not complicated. Put real people on your site. Write case studies that are specific about problems, approaches, and outcomes. Get some external citations. If you speak at events or publish in trade press, link to it.

Your HTML is a mess.

If your site was built on a page builder – Elementor, Divi, WPBakery, Visual Composer, or similar – the HTML it produces is genuinely terrible for AI crawling.

Page builders produce HTML that is wrapped in layers of div tags, inline styling, and presentational markup. The actual content – the stuff an AI crawler is looking for – is buried inside a structural mess that makes it harder to extract meaning from.

A hand-coded, semantic HTML site with properly structured headings, clean paragraph tags, and meaningful markup is far easier for a crawler to parse. The content layer and the presentation layer are properly separated. An H1 is the main heading. H2 and H3 tags reflect a genuine content hierarchy. Paragraphs contain content, not styling classes.

This is not a marginal difference. When an AI model is processing thousands of tokens from your page, the signal-to-noise ratio of your HTML matters. Clean code gives you more signal per token. A Divi-built page is a lot of noise.

You are not mentioned anywhere else.

AI tools do not just read your own website. They read everything.

If your business is mentioned on other credible websites – in articles, directories, case studies, partner pages, trade publications, review platforms, or professional associations – that mention contributes to the picture the AI has of who you are. Multiple consistent mentions of your business, services, and location from credible external sources tell an AI model that you are a real, established business in a particular field.

If the only place your business exists on the web is your own website, that is a thin evidence base. You might be brilliant at what you do, but from an AI’s perspective, you are unverified.

This is why PR, trade press coverage, directory listings on credible platforms, guest articles, and active profiles on LinkedIn and other platforms genuinely matter. They are not just vanity. They are citations – and citations build the picture that gets you recommended.

Your content is too thin or too vague.

Short, vague pages do not help AI tools understand what you do or who you serve.

A service page that says “we offer training and development solutions to organisations of all sizes” is not useful to anyone – human or machine. It does not say what sector you specialise in, what the training covers, what the outcomes are, or why you are better placed to deliver it than the next firm.

Depth matters. Not padding – actual depth. A service page that explains your methodology, describes the types of clients you work with, gives examples of outcomes, and answers the questions a buyer would have before picking up the phone is a far stronger candidate for an AI citation than a paragraph of marketing fluff.

This does not mean writing 3,000 words for the sake of it. It means making sure every page contains enough genuine information that someone – or something – reading it comes away with a clear understanding of what you do and why you are worth calling.

You have no schema markup.

Schema markup is structured data you add to your HTML that explicitly tells search engines and AI crawlers what type of content they are looking at. It is not magic, but it removes ambiguity.

For a B2B website, the most useful schema types include Organisation (your name, address, contact details, description), LocalBusiness if you serve a geographic area, Service to define individual service offerings, and Person to describe the people in your team. FAQ schema can also help if you have clear question-and-answer content on your pages.

Without schema, crawlers have to infer everything from your content. With schema, you are telling them directly. On a site where the content is already good, schema markup can push your signal clarity high enough to get you into the consideration set for AI recommendations that your competitors – who have no schema – are missing from.

Your site is too slow.

Page speed still matters here. Crawlers have budgets – they allocate a finite amount of time to crawling a site. A slow site means fewer pages crawled per visit. Fewer pages crawled means less content indexed. Less content indexed means less material for AI tools to draw from when building their picture of who you are.

A site that scores below 50 on Google’s PageSpeed Insights mobile test is not just a bad experience for human visitors. It is actively reducing the amount of your content that gets seen by crawlers. If your service pages are being partially crawled or skipped, the AI does not know they exist.

The same page builder sites that produce noisy HTML also tend to load slowly. The two problems compound each other.

What to do about it.

None of this is insurmountable. But it does require treating your website as a serious piece of infrastructure rather than a digital brochure you update once every few years.

The businesses that get mentioned in AI recommendations are not necessarily the biggest or the most established. They are the ones whose websites are technically sound, whose content is clear and specific, and whose presence across the web gives AI tools enough material to build a confident picture of who they are and what they do.

Start with a technical audit. Make sure your site is crawlable, fast, and built on clean HTML for AI optimisation. Then look at your content and ask whether it actually answers the questions your buyers are asking – not whether it sounds impressive, but whether it is genuinely useful. Then look at your external presence and ask whether there is enough third-party evidence to back up what your own site claims.

If the answer to any of those is no, that is where to start.