A recent study by the research firm Searchable has uncovered a significant digital disadvantage for smaller companies, revealing that 93% of London-based small and medium-sized enterprises (SMEs) are represented inaccurately by major large language models (LLMs). By testing platforms like ChatGPT, Perplexity, and Gemini over 13,000 times, researchers compared the visibility of SMEs against larger corporations. The findings confirmed a stark disparity: while 32% of large companies encountered inaccuracies, a staggering 50% of SMEs were subjected to at least one false, missing, or incomplete fact, highlighting a 56% higher rate of misinformation for smaller entities.
The discrepancy in data quality extends beyond mere omission, reaching into the realm of outright fabrication. According to the report, LLMs were twice as likely to “hallucinate” information about an SME compared to a larger brand, with a 5% fabrication rate. Furthermore, the analysis identified a persistent issue with brand identity; LLMs confused or misattributed the names of smaller businesses at a rate of 4%, nearly six times more frequently than they did for larger, more established organizations. These errors often centered on critical data points that potential customers rely on to make purchasing decisions, such as website URLs, phone numbers, founding dates, and core service offerings.
Industry experts believe this trend stems from the fundamental way LLMs are trained—relying on the sheer volume of web data to “learn” about the world. Because larger companies inherently command broader media coverage, more backlinks, and a more robust digital footprint, they are consistently prioritized by these systems. Smaller businesses, which lack the massive online visibility of global brands, effectively fall into a “discovery gap,” where AI models are either unable to find reliable information or mistakenly fill the void with incorrect “hallucinated” details.
This reliance on LLMs as a primary search and recommendation tool creates a tangible economic threat for smaller firms. Chris Donnelly, co-founder of Searchable, warns that when an AI provides a potential customer with a wrong phone number or an outdated list of services, it results in lost revenue and missed opportunities. By failing to accurately reflect the existence or capabilities of an SME, these AI systems act less like helpful assistants and more like barriers to entry, further exacerbating the existing divide between market giants and local businesses.
Despite these grim statistics, the study suggests that the landscape of AI-driven discovery is not necessarily as rigid as traditional search engine optimization (SEO). Unlike traditional search rankings, which often serve to reinforce the dominance of incumbent brands, AI-based retrieval systems operate on different logic. Donnelly argues that the playing field could potentially be leveled for smaller businesses if they take a proactive approach to their digital representation, ensuring their facts are distinct, verifiable, and optimized for AI-crawling technology.
Ultimately, the findings underscore the growing pains of a technological transition where consumers are increasingly turning to generative AI for quick, reliable answers. While LLMs are heralded for their intelligence, the inconsistency in their performance regarding local business data presents a significant challenge for the SME sector. Until AI developers refine their models to better handle lower-volume, hyper-local data, businesses must navigate a digital environment where their reputation and accessibility depend heavily on how accurately they are being “read” by a machine.
