SiteSignal
Research Reports
The State of AI Visibility in Sri Lankan Banking
Which banks ChatGPT, Gemini and Perplexity name and cite when people ask about banking in Sri Lanka, and how accurate their facts are.
SITESIGNAL
AI Visibility Research · November 2026
How AI answers UK workwear buyers: who gets named, and what gets cited
We asked ChatGPT, Perplexity, Gemini and Google AI Overviews the same 8 neutral UK workwear buyer questions every day for ten weeks. Here is who they named, what they cited, and what that means for a supplier.
8Neutral buyer questions
1,254Usable answers, 22 Aug to 1 Nov
17,409Citations
60Supplier sites checked
00Summary
The short version
We asked ChatGPT, Perplexity, Gemini and Google AI Overviews the same 8 neutral UK workwear buyer questions every day from 22 August to 1 November 2026. That gave 1,254 usable answers and 17,409 citations.
When a buyer asks “who”, AI almost always names suppliers, about 7 per answer. When a buyer asks “how” or “what should I look for”, AI almost never names one. About half of all citations go to supplier websites, and about a fifth go to government and public-sector pages. Reddit matters, but only in ChatGPT. Review sites, YouTube and “best of” lists hardly appear at all.
Website basics are not what separates the named suppliers from the rest. None of the 60 most-named supplier sites blocks AI crawlers in robots.txt, and sites with cleaner technical setups were not named more often.
Read the numbers on these questions
These are 8 questions and 2 SiteSignal accounts (2 UK workwear brands), so treat the numbers as a clear picture of these questions, not a census of the market. Ranges in brackets are 95% intervals. See How we measured.
01Who gets named
Ask “who” and AI names suppliers. Ask “how” and it almost never does.
On our 3 supplier-seeking questions, 89% of answers named at least one business (95% CI 70% to 99%; 3 questions, 520 answers). On the 5 advice questions, only 6% did (CI 2% to 9%; 5 questions, 734 answers).
Answers that named at least one business
What it means
If your buyers ask “how do I choose”, AI will teach them, not refer them. Being the source behind that advice is the main way in.
02Shortlist size
A supplier answer names about 7 businesses, and Gemini names the most
Answers to supplier-seeking questions named 7.0 businesses on average (CI 4.9 to 8.3). By engine: Gemini 11.0, ChatGPT 6.6, Perplexity 4.9, Google AI Overviews 2.9.
Businesses named per answer, supplier-seeking questions
Bar length is relative to Gemini’s 11.0.
What it means
AI gives buyers a shortlist, not one name. Google’s AI Overview gives the shortest list, so it is the hardest one to get onto.
03Citations
Half of what AI cites is other suppliers’ websites. A fifth is government.
Of 17,409 citations, 51% pointed to workwear supplier and other business sites (CI 36% to 62%). 19% pointed to government and public-sector pages such as gov.uk, hse.gov.uk and the public tender site (CI 9% to 34%). The 2 brands we track for were cited in only 1% (CI 0.1% to 2.7%) on these neutral questions.
Share of 17,409 citations
What it means
On neutral questions, AI leans on supplier sites and official guidance (safety rules, PPE, public tenders). Your site has to compete with both.
04Reddit
Reddit is a ChatGPT habit, not an AI habit
Reddit made up 17% of ChatGPT’s citations (CI 11% to 22%; 8,839 citations). In Perplexity it was 0.1% and in Google AI Overviews 0%. Wikipedia shows the same pattern: 13% in ChatGPT and almost none elsewhere.
Reddit’s share of each engine’s citations
Gemini is not shown: it ran without web search, so it gives no citations.
What it means
Each engine reads different sources. Look at the engine your buyers use before spending time on Reddit.
05Platforms
Review sites, YouTube and LinkedIn barely register
Review sites made up 0.0% of citations (CI 0.0% to 0.1%), YouTube 0.2%, LinkedIn 0.8% (CI 0.4% to 1.3%) and directories 1.0%, out of 17,409 citations on neutral questions.
| Source | Share of 17,409 citations |
|---|---|
| Directories and listings | 1.0% |
| 0.8% | |
| YouTube | 0.2% |
| Review sites | 0.0% |
What it means
For B2B workwear questions, AI isn’t reading reviews or videos. It reads guidance, supplier pages and forums.
06Page types
AI cites guides and official documents, not “best of” lists or product pages
Guides, blogs and advice articles made up 14% of cited pages (CI 6% to 28%). Government guidance pages were 13.5%, PDFs and official documents 13.4%, and homepages 13%. “Best X” listicles and comparison pages were only 1.0% (CI 0.4% to 1.9%), and product or category pages 0.5%. In Google AI Overviews, guides were 49% of citations.
Share of cited pages, by page type
Bars are on a 0 to 100% scale. The type comes from each web address, and about 1 in 5 pages could not be typed.
What it means
Write the useful guide (specs, standards, how to run a uniform tender). A product page or a “top 10” list is rarely what gets cited.
07Website basics
The most-named suppliers don’t block AI, and a cleaner site didn’t mean more mentions
We checked the 60 UK uniform websites AI names most. 0 of 60 blocked AI crawlers in robots.txt (CI 0% to 6%). 9 of 60 didn’t load for our checker (CI 8% to 26%). 20 of 60 had an llms.txt file. A simple readiness score (0 to 6) was not linked to how often a site was named (rank correlation -0.13, CI -0.39 to 0.14; 60 sites).
| Check on the 60 most-named supplier sites (15 October 2026) | Sites | 95% CI |
|---|---|---|
| Blocked AI crawlers in robots.txt | 0 of 60 | 0% to 6% |
| Didn’t load for our checker | 9 of 60 | 8% to 26% |
| Had an llms.txt file | 20 of 60 |
What it means
Blocking AI is rare in this industry and worth checking, because it is a quick fix. But a tidy site is not why suppliers get named. What other sites say about them matters more.
08Fact checks
AI says “I don’t know” more often than it gets a fact wrong
For the 2 brands we track for, from 18 September to 1 November 2026: 60% of fact checks were right (CI 45% to 73%), 36% got no specific answer (CI 24% to 51%), and 3.5% were wrong (CI 0.1% to 9.8%). That is 1,332 checks covering 9 facts per brand across ChatGPT, Perplexity and Gemini. Founder names and privacy policy links were the facts most often unknown. Founding year was the fact most often wrong.
1,332 fact checks on 2 brands, 18 Sep to 1 Nov 2026
Caveat: this covers only 2 brands. We use checks from 18 September only, because our fact-check scoring changed in mid-September, and the weekly numbers are still not steady. Treat this as a picture of these 2 brands, not a rate for the industry.
What it means
AI more often says it doesn’t know a company’s basic facts than gets them wrong. Make those facts easy to find.
09Method
How we measured
- Questions: 8 neutral buyer questions that don’t name any brand: 3 supplier-seeking (“which suppliers...”, “best options for ordering...”, “who helps write a uniform specification...”) and 5 advice questions. They come from 2 UK workwear brands’ SiteSignal accounts. We left out 4 more questions because they named a brand.
- Engines: ChatGPT (OpenAI API with web search), Perplexity (API), Gemini (API, without web search, so Gemini gives no citations) and Google AI Overviews (via Google search results).
- Dates: 22 August to 1 November 2026. All 8 questions ran to 26 September. 4 ran to 1 November. Each question was asked once a day in each engine.
- Answers: 1,254 usable answers. We dropped Google checks where no AI Overview appeared or the request failed, and kept 95 AI Overviews. There were 17,409 citations from ChatGPT, Perplexity and AI Overviews.
- Uncertainty: asking the same question every day gives very similar answers, so we treat the question, not the answer, as the unit. Ranges are 95% intervals from resampling questions (2,000 resamples). The site check uses a standard interval for a proportion.
- Site check: one check of the 60 most-named supplier homepages on 15 October 2026: page load, robots.txt, AI bot user agents, llms.txt and structured data.
- Limits: this is 8 questions in one industry, from 2 accounts. That is below our own bar of 10 questions for a headline number, so read the figures as “on these questions”. Citation classes come from each web address, and about 1 in 5 pages could not be typed. “Businesses named” counts every brand or organisation named in an answer.
The 2 brands are kept anonymous, and so are the private businesses AI cited. Only large public platforms and public-sector sites are named.
10Data
Download the data
Aggregate tables behind each finding, as CSV files. They hold totals and percentages only, with no answer text and no private business names.
- f1_answers_naming_brands.csv Findings 1 and 2, by engine and question type
- f2_citation_mix.csv Findings 3 to 5, citation sources by engine
- f2_top15_domains.csv Top 15 cited domains, private businesses not named
- f3_page_types.csv Finding 6, cited page types by engine
- f5_fact_checks.csv Finding 8, fact check results by engine
11Questions
Frequently asked questions
Does AI name workwear suppliers when buyers ask for advice?
Almost never, on these questions. On 3 supplier-seeking questions, 89% of answers named at least one business (95% CI 70% to 99%; 520 answers). On 5 advice questions, only 6% did (CI 2% to 9%; 734 answers).
How many suppliers does an AI answer name?
Answers to supplier-seeking questions named 7.0 businesses on average (CI 4.9 to 8.3). By engine: Gemini 11.0, ChatGPT 6.6, Perplexity 4.9, Google AI Overviews 2.9.
What sources does AI cite for UK workwear questions?
Of 17,409 citations, 51% pointed to workwear supplier and other business sites (CI 36% to 62%) and 19% to government and public-sector pages such as gov.uk and hse.gov.uk (CI 9% to 34%). Review sites were 0.0%, YouTube 0.2% and LinkedIn 0.8%.
Is Reddit important for AI visibility in workwear?
Only in ChatGPT. Reddit made up 17% of ChatGPT’s citations (CI 11% to 22%; 8,839 citations), 0.1% in Perplexity and 0% in Google AI Overviews.
Do the most-named workwear suppliers block AI crawlers?
No. 0 of the 60 UK uniform websites AI names most blocked AI crawlers in robots.txt (CI 0% to 6%). A simple readiness score was not linked to how often a site was named (rank correlation -0.13, CI -0.39 to 0.14; 60 sites).
Your AI search strategist
See what AI says about your brand, then fix it
SiteSignal checks ChatGPT, Gemini, Perplexity and Google AI Overviews for you and turns the gaps into a weekly to-do list.
No credit card needed.