Most UK B2B tech sites miss AI readability signals

New research from Marketing Graham found that 77.5% of UK B2B technology websites are missing at least one basic AI machine-readability signal. The audit covered 752 UK tech companies.

The findings point to a gap in how technology suppliers present themselves to automated tools increasingly used by buyers to research software vendors and service providers. When those signals are absent, AI assistants can struggle to find, read and interpret a company's offering.

Of the 752 websites reviewed, 583 lacked at least one of four recognised markers: structured data, an XML sitemap, a readable robots.txt file or llms.txt. Only 22.5% of the audited websites had all four in place.

The research also suggests the shortfall is often narrow rather than systemic. Of the 583 websites that fell short, 325 were missing only llms.txt, a text format intended to give large language models a clear summary of what a company does.

More established website signals were in a stronger position. The data showed that 90.3% of sites had robots.txt, 87.6% used XML sitemaps and 69.3% had structured data.

Discovery shift

The report comes as software buyers increasingly use AI tools such as ChatGPT, Claude and Perplexity alongside conventional search engines when drawing up supplier lists. For B2B technology companies, that raises the importance of machine-readable website information, because AI systems may rely on a mix of direct site crawling, scraped pages and third-party references.

The issue affects large and small businesses at similar rates, suggesting established enterprise brands have no clear advantage over younger scaleups in basic AI discoverability. That may create an opening for smaller firms that move quickly to address the missing signals.

Graham Smith, Fractional Marketing Director at Marketing Graham, led the research. "You would expect B2B tech firms to be leading from the front when it comes to AI readiness, but the findings show a surprising gap between what the industry preaches and how it manages its own digital housekeeping.

Despite that, more than half of the companies falling short already have three of the four key signals, which suggests they don't need to start from scratch. Much of the established technical infrastructure is already there. It's just the newer llms.txt signal that needs to be incorporated," Smith said.

Technical baseline

The index measures four publicly accessible website elements: Schema.org markup using JSON-LD, Microdata or RDFa; a valid XML sitemap; a readable robots.txt file; and a genuine, non-empty llms.txt resource.

The consultancy does not argue that each signal carries the same weight or that having all four guarantees appearance in AI-generated answers. Instead, it presents the list as a set of observable technical indicators that may help machines discover, crawl or interpret site information.

That distinction matters because AI search visibility is shaped by more than website files alone. Relevance, authority, useful content and external citations can all influence whether an AI assistant mentions a company in response to a buyer query.

Smith said those broader factors still depend on a basic level of technical accessibility. "Authority, useful content, third-party citations, relevance and clear information about who you are all matter if you want to be recommended by an AI assistant, but it's really important to ensure that machines can also read the signals in the first place," he said.

The low adoption of llms.txt stands out because it contrasts with the broader strength of conventional search infrastructure across the UK tech sector. While most companies appear to have the foundations for search engine indexing, far fewer have updated their sites for newer AI-focused conventions.

Smith said the industry should be careful about dismissing the newer file format simply because it has not yet become standard. "The research shows that's where many tech firms are falling short and, as a result, struggling to be seen. Emerging signals such as llms.txt could give early adopters, including smaller companies, an advantage.

While the majority of websites scored well on the more established signals, llms.txt was present on only a quarter of the sites analysed. But we have to acknowledge that it was only proposed in 2024. Robots.txt took 28 years to be formalised as an industry standard.

Nobody knows whether llms.txt will become as established. However, businesses should be cautious about overlooking or dismissing it. In many cases, implementing it is as easy as flicking a switch. It doesn't interfere with the website customers see, and LLMs that don't use it can simply ignore it," he said.

The data suggests many UK tech companies may not need major website redevelopment to improve AI visibility. In many cases, the gap appears to come down to one missing file rather than a broader failure of site structure or indexing practice.

Only a fifth of the websites audited currently show all four signals, while more than three quarters are missing at least one.