How citable are the UK’s fastest-growing tech companies to AI search?

01The median Fast 50 company scores 67 out of 100

Across the 47 companies we scored, the median score is 67 and the mean is also 67. The middle half sits between 64 and 71. We crawled up to 40 pages on each site, 1,783 pages across the 47 scored sites, and ran every page through Rubric’s checks (44 in this scoring version, with the broken-link check switched off). A site’s score is the average of its pages, and we only use the median to compare sites.

Thirty-three of the 47, seven in ten, score below 70. In Rubric’s benchmark of 244 websites, crawled on 7 October 2026, 77% did. So a slightly bigger share of the Fast 50 clears the line. The benchmark covers online shops, blogs and publishers, SaaS companies and other businesses. The medians, 67 and 66, are almost level, and the samples differ, so that’s as far as the comparison goes.

In GoGoChimp’s 11 October 2026 crawl, seven in ten of the Deloitte UK Technology Fast 50 2025 companies we scored, 33 of 47, fell below Rubric’s quotable line of 70.

Split by Deloitte’s own sector labels, the 20 B2B or B2B2C software companies have a median of 67.5 and the 16 fintechs 66. That’s a small gap for groups this size. The environmental technology and B2C software groups have four companies each, and that’s too few to compare.

02Trusted is the weakest pillar, with a median of 47

Twenty-nine of the 47 companies score under 50 on Trusted, and only three reach 70. Rubric sorts its checks into three pillars and gives each its own score. Trusted is the one that asks for evidence: named authors, links to outside sources, sourced statistics and freshness.

Findable edges Known as the strongest pillar. Most of these sites use one H1 per page and put their main answer where a crawler can read it without running JavaScript. They lose their Findable points on openers and headings, and both turn up in the failing checks below.

Trusted is the pillar where the Fast 50 companies score lowest: a median of 47, against 72 for Known and 73 for Findable.

03Which checks do Fast 50 websites fail most often?

Named authors fail most often, at 41 of the 47 companies, then content-type schema at 36. Six of Rubric’s checks fail at 25 or more of the 47. A company fails a check here when it comes back bad on more than half its applicable pages. Those are the pages the check runs on, and some checks don’t run on every page.

Named authors are the clearest gap. Rubric treats a named author as a trust signal: a real person who stands behind the page and can be checked.

The 40 to 70 word answer-first opener is our house standard, and no engine requires it. The check asks whether a page opens with a self-contained answer that could be quoted on its own. Our guide to content grounding in AI search shows how to write sentences an AI answer can rest on.

None of the 47 failed Rubric’s HTTP, indexability, canonical, robots.txt, internal linking or alt-text checks on most of their pages.

In GoGoChimp’s 11 October 2026 crawl, 41 of the 47 Deloitte UK Technology Fast 50 2025 companies we scored failed Rubric’s named-author check on more than half the pages it applied to.

04Blocking AI training bots doesn’t hide a site from AI search

Blocking AI training crawlers doesn’t block the search bots, because the vendors run the two as separate bots. Lottie (#3 on the Fast 50) blocks six of the 15 AI crawlers Rubric checks in robots.txt, all of them training or data crawlers, GPTBot and ClaudeBot among them. It lets through the bots that fetch pages for AI search answers, including OAI-SearchBot (OpenAI), PerplexityBot (Perplexity) and Googlebot (Google). It scores 73, joint sixth on the list.

OpenAI and Anthropic both describe GPTBot and ClaudeBot as training crawlers (OpenAI, Anthropic). Of the 15 AI crawlers Rubric checks, Lottie was the only company on the list to block any outright in robots.txt. It’s the split to copy if you want your content kept out of model training but still reachable by AI search. Block the training crawlers by name and leave the search bots alone.

In GoGoChimp’s 11 October 2026 crawl of the Deloitte UK Technology Fast 50 2025, Lottie was the only company to block any of the 15 AI crawlers Rubric checks outright in robots.txt, and it blocked training crawlers such as GPTBot and ClaudeBot while letting the AI search bots through.

We also sent the sites test requests from our own IP address, labelled with AI and search bots’ user agents. One site refused all six AI and search bot user agents we sent, Googlebot included. Another refused only Googlebot and Bingbot, which looks like a firewall turning away search bots it can’t verify. A third refused only ClaudeBot, Anthropic’s training crawler. The engines’ own verified addresses may get a different response, so treat these as signals to check. None of them proves a block.

05Homepages outscore the rest of the site at 45 of 47 companies

At 45 of the 47 companies, the homepage beats the site’s median inner page, by a median of 13 points. The median homepage scores 80 and the median inner page 67. A check that stops at the homepage flatters almost every site on the list.

My read is that the homepage is the page everyone fusses over. A buyer with a specific question usually lands on an inner page, like a product page or a guide. That’s where these sites score lower. We’ve seen the same pattern in our benchmark of SEO agency websites.

In GoGoChimp’s 11 October 2026 crawl, the homepage outscored the median inner page at 45 of the 47 Deloitte UK Technology Fast 50 2025 companies we scored, by a median of 13 points on Rubric.

06Some sites serve schema that crawlers without JavaScript can’t see

In our second crawl, two of the 47 companies had schema that only appeared once JavaScript ran, on 2 and 6 pages. Earlier the same night, in the first crawl, three companies showed it, on 2, 13 and 15 pages. Rubric spots it by comparing each page’s raw HTML with a rendered copy.

Vercel’s December 2024 analysis found that “none of the major AI crawlers currently render JavaScript”, OpenAI’s crawlers, ClaudeBot and PerplexityBot included (Vercel and MERJ, 2024). A crawler that doesn’t run JavaScript never sees schema a script adds. Our guide to which AI crawlers to allow, and whether they render JavaScript goes further on the crawler side.

Rubric catches JavaScript-only schema by comparing each page’s raw HTML with a Chrome-rendered copy. Across all 415 agencies in the crawl, 44 (10.6%) had at least one page like this. At 1 of 42 (2.4%), the North is cleaner than the wider benchmark here. If you’re not sure where your own markup sits, check whether your schema needs JavaScript before assuming the AI crawlers see it.

Eight of the 47 don’t have schema on any page we crawled.

07Twenty-four of the 47 companies declare no sameAs links

Of the 47 companies, 24 don’t declare any sameAs links, and two link a Wikidata entry. A sameAs link sits in Organization schema. It declares that a profile elsewhere, such as the company’s LinkedIn or Crunchbase page, belongs to the same company. Rubric’s entity-clarity check looks for it.

Sixteen of the 47 publish an llms.txt file, a proposed plain-text summary of a site written for AI tools. Rubric checks for it but doesn’t score it.

08Most of the Fast 50’s dated articles are more than a year old

Of the 169 dated articles we crawled, 87 are more than a year old, and the median one is 370 days old. Rubric’s crawler labelled 197 pages across the 47 sites as articles, such as blog posts and guides. Of those, 169 carry a date. Because the crawl stops at 40 pages, a site’s newest posts aren’t always in it.

Seer Interactive’s June 2025 log study looked at three ChatGPT bots. Nearly 65% of their hits were on content published within the past year (Seer Interactive, 2025). That’s a count of what the bots fetched. It says nothing about what ChatGPT cited, and it’s correlational. It backs Rubric’s freshness check without proving that engines prefer newer pages.

09Which Fast 50 companies scored highest, and why?

Summize leads on 78. City Unscripted and Econic Technologies follow on 75, and Under The Doormat and ValStream are level on 74. Econic has the highest Trusted score in the sample and City Unscripted the second highest. The scores reflect page markup and structure on the crawl date. They don’t judge the companies’ products.

Summize (78, #10 on the Fast 50) has the sample’s highest Known score, 91. It links to outside sources on 34 of the 38 inner pages where that check applied.

City Unscripted (75, #5) scores 81 on Trusted. None of the schema on its 39 inner pages needs JavaScript to appear.

Econic Technologies (75, #24) scores 84 on Trusted. It names an author on 29 of the 37 inner pages where that check applied, and links to outside sources on all 37.

The lowest four scores are 57, 59, 60 and 60, and we haven’t named those companies.

10Why weren’t three of the companies scored?

Two of the three served our crawler a firewall challenge page on most of their URLs. On the third, only the homepage was crawlable. We leave a site unscored when most of its pages are challenge pages or the crawl can’t get past the homepage. Neither gives a fair read of the site.

There’s nothing on a challenge page for AI to quote. From outside, we can’t see whether verified AI crawlers get the same page our crawler did. If you run one of the challenged sites, check what your firewall serves the AI search bots.

11How did we crawl and score the sites?

We crawled all 50 company websites twice on 11 October 2026 with Rubric’s audit engine, at up to 40 pages a site. We scored 47 of them. The crawls ran about an hour apart. Every figure here comes from the second, which ran on a newer scoring version, except where we compare the two crawls.

• Sample. The Deloitte UK Technology Fast 50 2025, published 7 November 2025, which ranks UK technology companies by average revenue growth over a three-year period (Deloitte, 7 November 2025). It was the latest edition when we crawled. It represents fast-growing UK tech companies, so don’t read it as UK business in general.
• Domains. We used the website Deloitte links for each company. For the six it doesn’t link, we confirmed the domain by fetching it and checking the page title.
• Crawl. We ran Rubric’s audit engine locally, outside the hosted Rubric service, with the broken-link check off. It reached 8 pages on one scored site, 16 on another and 25 or more on the rest.
• Scoring. Rubric scoring version 2026-10-11: 44 checks across three pillars (Known, Findable, Trusted), rolled into a 0 to 100 score. A site’s score is the average of its pages. We use the median only to compare sites.
• Two crawls. The second crawl’s scoring version (2026-10-11) corrected how Rubric treats a robots.txt block on AI training crawlers. Before the fix, Rubric counted that as a block on AI search and capped the score. That’s why Lottie couldn’t be scored in the first crawl.
• Benchmark. The 244-site comparison is Rubric’s benchmark corpus, re-crawled on 7 October 2026 at the same 40-page cap on the previous scoring version (2026-10-01). On the 44 Fast 50 sites scored under both versions, the version change barely moved the scores. The same fix does change the benchmark’s make-up, though. Eleven sites it left out on 7 October, because the old rule capped their robots.txt blocks on AI training crawlers, would now be scored, at 54 to 68 with a median of 62. So its median may slip by about a point once the whole corpus is re-crawled. It’s a different sample from the Fast 50, so treat the comparison as context only.
• What the score is not. Rubric estimates citability from on-page, structural and technical signals. It doesn’t count citations a company has already earned, and it doesn’t judge the company’s product.

How stable are the scores?

The scores barely moved between the two crawls. For the 44 companies scored both times, the median company moved by 1 point, up or down, and none moved by more than 3. The median of the 44 went from 67 to 67.5. The scoring version changed between the crawls too, so the result covers that change as well.

Across GoGoChimp’s two 11 October 2026 crawls of the Deloitte UK Technology Fast 50 2025, none of the 44 companies scored both times moved by more than 3 points on Rubric.

12How do you check your own site for the same gaps?

Start with the inner pages, because that’s where these sites scored lower. Pick a product or service page and a guide. Check each for the gaps this crawl found most often: a named author, content-type schema and an answer-first opener. You’ll need a browser and a few minutes a page, plus a terminal for the last step.

1. Look for a byline with a real name, then search the page source for “author”.
2. Type view-source: before the URL in Chrome and search for application/ld+json. You’re looking for a type that says what the page is, such as Article on a guide or Service on a service page.
3. Open the same page in the browser inspector and search again. Schema that shows there but not in view-source is added by JavaScript, so a crawler that doesn’t run scripts won’t see it.
4. Read the headings on their own and mark each as a question, a claim or a label. Then read the first 70 words under each and ask whether they’d make sense quoted alone.
5. Count the external links in the body copy. Two is Rubric’s floor.
6. Search the homepage source for “sameAs” and count the profiles it lists.
7. Open /robots.txt and see which bots it names. A block on GPTBot or ClaudeBot opts out of training. OAI-SearchBot (OpenAI) and PerplexityBot (Perplexity) are the search bots to leave alone.
8. Request a page as a bot, such as curl -A "OAI-SearchBot" https://www.example.com/, and see whether you get the page or a challenge. A refusal isn’t proof the real bot is blocked, because your request doesn’t come from its network.

If you’d rather not do it by hand, run a free Rubric audit on your own site. There’s no sign-up, and it takes about a minute. Compare its failing checks with the six in the table above. Don’t set its headline score against ours. Your audit may read a different number of pages from our 40, and that alone can shift the score.

References

• Per-site scores: not published. Any company on the list can request its own results.
• Rubric benchmark corpus, re-crawled 7 October 2026 (scoring version 2026-10-01): 244 scored sites, median 66, 77% below 70, Trusted median 49. Published as GoGoChimp’s state of AI citability study: https://www.gogochimp.com/research/state-of-ai-citability-2026
• Google Search Central (undated). AI features and your website. https://developers.google.com/search/docs/appearance/ai-features
• Seer Interactive (2025, 25 June). Study: AI brand visibility and content recency. https://seerinteractive.com/insights/study-ai-brand-visibility-and-content-recency
• Deloitte (2025). UK Technology Fast 50 2025 winners. https://www.deloitte.co.uk/fast50/winners/2025/
• Vercel with MERJ (2024, 17 December). The rise of the AI crawler. https://vercel.com/blog/the-rise-of-the-ai-crawler
• Anthropic (2026, 7 April). Does Anthropic crawl data from the web, and how can site owners block the crawler? Anthropic Help Center. https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler
• Deloitte (2025, 7 November). Journee tops Deloitte UK Technology Fast 50 ranking (press release). https://www.deloitte.com/uk/en/about/press-room/journee-tops-deloitte-uk-technology-fast-50-ranking.html
• OpenAI (undated). Overview of OpenAI crawlers. https://developers.openai.com/api/docs/bots
• Perplexity (undated). Perplexity crawlers. https://docs.perplexity.ai/guides/bots