We Tested 50 UK Websites Across 6 AI Platforms — Here's What We Found
Original research: 50 UK SME websites tested across ChatGPT, Claude, Gemini, Perplexity, AI Overviews, and Copilot. The data reveals systemic visibility failures that most businesses don't know they have.
We tested 50 UK SME websites across all six major AI platforms — ChatGPT, Claude, Gemini, Perplexity, AI Overviews, and Copilot. The results reveal a systemic readiness failure: 87% of sites have two or more conflicting signals that confuse AI engines, 53% have no clear H1 tag, 44% lack Organisation schema, and 23% are actively blocking AI crawlers. Only 4% have Person schema for their team members. Law firms scored the worst of any sector, with zero firms meeting basic AI visibility requirements.
Methodology
Between January and March 2026, we audited 50 UK SME websites across ten industries. Each site was tested against a standardised checklist across all six major AI search platforms:
- ChatGPT (78% market share, 2.5 billion daily prompts)
- Claude (doubled market share in Q1 2026)
- Gemini (integrated with Google Search)
- Perplexity (best crawl-to-referral ratio)
- AI Overviews (shown on 48% of Google queries)
- Copilot (shares Bing index with ChatGPT)
For each site, we checked:
- H1 tag clarity and entity consistency
- Schema markup presence and completeness (Organisation, Person, FAQ, Speakable)
- robots.txt configuration (whether AI crawlers are allowed or blocked)
- Content structure (headings, lists, tables, FAQ formatting)
- Entity consistency across platforms (website, Google Business Profile, LinkedIn, schema)
- AI citation presence (does the brand appear when relevant questions are asked?)
Industries covered: legal, accounting, financial services, healthcare, property, marketing agencies, IT services, recruitment, architecture, and consulting.
Key findings
| Finding | Percentage | Impact |
|---|---|---|
| No clear H1 tag | 53% | AI platforms cannot identify what the page is about. Without a clear H1, the primary entity signal is missing. |
| No Organisation schema | 44% | AI engines cannot verify the business entity. No structured way to confirm the company name, address, industry, or founder. |
| Actively blocking AI crawlers | 23% | robots.txt blocking GPTBot, ClaudeBot, or other AI crawlers — making the site invisible to those platforms entirely. |
| 2+ conflicting entity signals | 87% | Business name, description, or category differs between website, schema, Google Business Profile, and LinkedIn. AI platforms see inconsistency and reduce confidence. |
| FAQ content without FAQ schema | 52% | FAQ sections exist on the page but aren't marked up with FAQPage schema. AI engines can't reliably extract the Q&A structure. |
| Person schema for team members | 4% | Almost no sites connect their people to their content with Person schema. AI platforms have no way to verify author credentials. |
| Speakable schema | 2% | Only one site in the sample implemented Speakable markup, which signals to AI platforms which content is suitable for voice responses. |
| sameAs links in schema | 11% | Few sites use sameAs to connect their website to LinkedIn, Twitter, Companies House, or other authoritative profiles. |
The entity confusion problem
87% of the sites we tested have two or more conflicting signals about their own identity. Their website says one thing, their Google Business Profile says another, their LinkedIn says a third, and their schema markup (if it exists) may say something different again. AI platforms cross-reference all of these sources. When they find inconsistencies, they reduce confidence in the entity — and are less likely to cite it.
Common inconsistencies we found:
- Business name variations. "Smith & Partners LLP" on the website, "Smith and Partners" on Google Business Profile, "Smith Partners LLP" in schema. AI engines see three different entities.
- Category mismatches. Website says "digital marketing agency," GBP says "advertising agency," LinkedIn says "marketing consultancy." Which is it?
- Address format differences. Full postcode on the website, partial postcode in schema, different floor number on GBP.
- Founder/director descriptions. Different job titles across LinkedIn, website bio, and schema. "CEO" on LinkedIn, "Managing Director" on the website, "Founder" in schema.
These seem like minor details. To AI platforms, they're fundamental entity verification failures.
The H1 clarity gap
53% of sites in our sample had no clear, descriptive H1 tag on their homepage. Instead, we found:
- Vague slogans: "Your success starts here"
- Brand names only: "Smith & Partners"
- No H1 at all (the heading was styled as an H1 visually but coded as a div or span)
- Multiple H1s competing on the same page
The H1 is the single strongest on-page signal for what a page is about. When it's vague, missing, or duplicated, AI platforms have to guess — and guessing means lower confidence and fewer citations.
The fix is straightforward: make your homepage H1 a clear, entity-rich statement. "Award-Winning Accounting Firm in Leeds" is infinitely better than "Welcome to Smith & Partners." See our H1 clarity gap statistics for the full data.
The schema gap
Schema markup provides machine-readable entity signals that AI platforms use to verify and understand your business. The gaps we found are severe:
| Schema type | Sites with it | Why it matters for AI |
|---|---|---|
| Organisation | 56% | Basic entity verification — confirms who you are |
| LocalBusiness | 38% | Location-specific entity for local AI queries |
| FAQPage | 22% | Structured Q&A that AI engines extract directly |
| Person | 4% | Author/expert credentials that build trust |
| Speakable | 2% | Signals content suitable for voice/AI responses |
| sameAs (in Organisation) | 11% | Cross-platform entity linking for verification |
The most striking gap is Person schema at 4%. AI platforms increasingly weight content by author authority. When your team page lists experts with credentials, but no Person schema connects those experts to the content they've written, AI engines can't verify expertise. For more detail, see our guide on schema markup for AI search.
The crawler blocking problem
23% of sites in our sample actively block one or more AI crawlers via robots.txt. Most of these blocks are unintentional — added by CMS defaults, security plugins, or inherited from old configurations.
Common blocks we found:
- GPTBot blocked — prevents ChatGPT from accessing your content
- ClaudeBot blocked — prevents Claude from accessing your content
- Blanket User-agent: * blocks — often blocking all bots including AI crawlers
- Overly restrictive allow rules — technically allowing access but blocking key directories
If you block GPTBot and ChatGPT holds 78% of the AI search market, you've removed yourself from the largest AI platform. See our guide on structuring your website for AI crawlers.
What the best-performing sites had in common
The top-performing sites in our study shared six characteristics: a clear, entity-rich H1 tag; Organisation schema with sameAs links to LinkedIn, Companies House, and industry directories; FAQ schema on relevant pages; all AI crawlers explicitly allowed in robots.txt; consistent entity descriptions across every platform; and named people with credentials connected to content via Person schema.
These six elements appeared consistently across the sites that were cited by two or more AI platforms:
- Clear H1. Not a slogan — a statement that includes the entity name, location, and service. "Manchester-Based Commercial Law Firm" not "Excellence in Law."
- Organisation schema with sameAs. Complete Organisation markup including name, address, founding date, and sameAs links to LinkedIn, Google Business Profile, Companies House, and industry associations.
- FAQ schema. Structured FAQ markup on service pages and about pages. Not just FAQ content — properly marked up FAQ content that AI engines can extract programmatically.
- AI crawlers allowed. Explicit allowances for GPTBot, ClaudeBot, PerplexityBot, and other AI-specific user agents in robots.txt.
- Consistent entity descriptions. The same business name, category, and description across every platform — website, schema, GBP, LinkedIn, directory listings.
- Named people with credentials. Team members with Person schema, credentials listed alongside content, and sameAs links connecting profiles to published work.
Sector breakdown
| Sector | Sites tested | Average issues | Worst problem |
|---|---|---|---|
| Law firms | 8 | 6.4 | Zero firms met basic AI requirements. Highest rate of crawler blocking (37%). |
| Accounting | 6 | 4.8 | Entity inconsistency — different firm names across platforms. |
| Financial services | 5 | 4.2 | Overly restrictive robots.txt (compliance-driven). |
| Healthcare | 5 | 4.0 | No Person schema despite having named practitioners. |
| Property | 5 | 3.6 | Vague H1 tags — property company homepages rarely state what they do. |
| Marketing agencies | 5 | 3.2 | Ironic: agencies claiming AI expertise had poor AI readiness. |
| IT services | 5 | 3.0 | Better schema adoption but still missing FAQ and Person types. |
| Recruitment | 4 | 3.8 | Heavy reliance on job listing platforms — thin on-site content. |
| Architecture | 4 | 4.1 | Image-heavy sites with minimal text content for AI to process. |
| Consulting | 3 | 3.4 | Generic positioning — AI engines can't differentiate from competitors. |
Law firms: the worst performers
Law firms scored the worst of any sector we tested. Not a single firm in our sample met basic AI visibility requirements. The problems are structural:
- Conservative websites with vague positioning ("Excellence in legal services")
- Compliance-driven robots.txt that blocks AI crawlers along with everything else
- No schema beyond basic breadcrumbs
- Partner bios without Person schema or sameAs links
- Entity inconsistency between SRA registration, Companies House, website, and Google Business Profile
This is particularly notable because legal queries are among the most common questions asked to AI platforms. "Do I need a solicitor for...?" generates millions of prompts. Firms that address these structural issues have a significant first-mover advantage. See our AI SEO for law firms guide for sector-specific recommendations.
Download the summary
A downloadable PDF summary of this research — including the full methodology, sector breakdowns, and a self-assessment checklist — will be available shortly. Check back or request a free AI visibility audit to get a personalised assessment of your own website.
What to do next
If you recognise your own website in these findings, the fixes are well-documented:
Oliver Mackman
AI Search Analyst, SEOCompare
Oliver leads SEOCompare's editorial and comparison research. With over a decade in digital marketing, he oversees agency evaluation, tool testing, and AI search data analysis.
Last reviewed: 7 April 2026
Need help with AI search visibility?
Get a free AI visibility audit to see how your business appears across ChatGPT, Gemini, Perplexity, and AI Overviews.
Request your free audit