SiteSignal
When people ask AI engines for a company’s facts, such as its contact details, founder or policy links, the engines are far more likely to say they don’t know than to give a wrong answer. That held in two unrelated sectors SiteSignal tracks: UK uniform and workwear suppliers, and Sri Lankan banking.
Key findings
- UK workwear: across 6,199 fact-checks on the brands we track (May to October 2026), 51% of answers were right, 10% partly right, 35% gave no specific answer, and about 5% were wrong.
- The engines behave differently: Gemini gave no specific answer 65% of the time. ChatGPT was right 70% of the time and wrong 3%. Perplexity was wrong 8.5% of the time and partly right 20%.
- Sri Lankan banking: in 1,272 fact-checks on one bank, AI gave no wrong email addresses. The only errors were an outdated phone number and a link to a page that no longer exists, in February and March 2026, and neither appeared from May.
- AI almost never cites criticism: 0.4% of 58,299 workwear citations and 0.2% of 34,532 banking citations carried negative sentiment.
What it means
AI engines often can’t find a company’s facts, so they say nothing. The fix is simple: put your contact details, policies and key facts on one clear page, keep them identical everywhere they appear, and retire old phone numbers and links.
About the data
SiteSignal checks what ChatGPT, Gemini and Perplexity say about a business against the business’s own answers. Answers of “Unknown” count as no answer. The workwear figures cover the brands tracked in two SiteSignal accounts in that sector; the banking figures cover one bank’s checks. Full method: UK Uniform & Workwear, Sri Lankan Banking and Workwear vs. Banking.
Press contact: hi@sitesignal.app