AI assistants confuse one in 20 companies with an entirely different business, new research finds

LLM Listed launches AI Brand Protection service after study finds leading AI platforms routinely omit, distort or provide inconsistent information about businesses

LONDON, UK — 13/08/2026 — Leading AI assistants confused 5% of companies with entirely different businesses and omitted important products or services in 28% of responses, according to new research examining how ChatGPT, Gemini, Claude and Perplexity represent 250 companies.

The State of AI Business Accuracy Report 2026, published by AI visibility company LLM Listed, evaluated companies across 20 industries to determine how accurately the leading AI platforms describe their identity, leadership, location, products, services, pricing and reputation.

The research found that although AI correctly identified the core purpose of a business in 92% of responses, its understanding frequently became less reliable when asked for commercially important details.

Only 46% of companies received broadly consistent recommendations across all four platforms. Local businesses were four times more likely than nationally recognised brands to be materially misunderstood.

Among the report’s principal findings:

- 28% of responses omitted important products, services or specialist capabilities.
- 5% confused the company with a different business.
- Only 46% of companies received consistent recommendations across all four platforms.
- Local businesses were four times more likely to be materially misunderstood than national brands.
- AI correctly identified company founders in 70% of responses.
- AI correctly identified company headquarters in 85% of responses.
- 23% of pricing answers included specific figures, many of which appeared speculative or could not be confidently linked to authoritative public sources.
- Companies’ own websites appeared as sources in 40% of recommendation responses.
- Reddit appeared in 35% of recommendation responses.

The results suggest that a company can be recognizable to AI while still being inaccurately represented at the point when a potential customer is evaluating it.

Ben Harper, founder of LLM Listed, said:
“Most companies are asking whether they appear in ChatGPT. We think there is now a more urgent question: is what ChatGPT says about the company actually true?

“An AI assistant can correctly understand the broad purpose of a business while getting an important detail wrong. It might omit a key service, quote an old price, name a former executive, misunderstand where the company operates or tell a customer that a product lacks a capability it already has.

“Those errors happen before the customer reaches the company’s website, so they are largely invisible to conventional analytics. The company may never know that it was eliminated from consideration because of inaccurate information.”

Incorrect information can affect purchasing decisions

AI assistants increasingly act as an intermediary between businesses and their customers.

People use them to identify potential providers, compare products, investigate reputations, assess suitability and understand pricing. An incomplete or outdated answer can therefore influence a purchasing decision without generating a measurable website visit.

The consequences can be particularly significant when the incorrect information concerns:

- Product capabilities and integrations
- Prices and commercial terms
- Geographic availability
- Customer suitability
- Accreditations and regulatory status
- Security and privacy credentials
- Company ownership
- Current leadership
- Professional services and specialist expertise

The report found meaningful differences between AI platforms. In 54% of companies tested, at least one platform materially diverged from the others in how the company was described or recommended.

Harper continued:
“There is no single definitive AI answer about a company. ChatGPT may represent it accurately while Gemini relies on older information, or Perplexity may emphasize a source that produces a completely different conclusion.

“That means checking one platform once is not enough. Companies need to understand whether important errors repeat, which questions trigger them and what evidence may be influencing the answer.”

LLM Listed launches AI Brand Protection

In response to the findings, LLM Listed has launched LLM Listed Brand Protection, a service designed to find and address false, outdated, misleading or commercially damaging information produced by AI platforms about companies.

The service tests how AI represents the facts most likely to influence customer trust and purchasing decisions. These can include products, pricing, locations, leadership, qualifications, security, integrations and suitability for particular customers or use cases.

Findings are assessed according to their accuracy, reproducibility, supporting evidence and potential commercial or reputational significance.

Where a material discrepancy is identified, LLM Listed investigates the information environment that may be contributing to the answer. This can include conflicting first-party pages, outdated directories, old company profiles, third-party comparisons, incomplete product documentation and insufficient authoritative evidence.

The company then works with the client to strengthen or correct the relevant information and repeat the original tests.

The initial AI Brand Protection Sprint is a fixed 30–60 day programme that includes:

- Creation of a verified company fact base
- Multi-platform accuracy testing
- Identification of repeatable inaccuracies
- Commercial and reputational severity assessment
- Investigation of conflicting and missing sources
- Prioritised correction work
- Retesting and resolution tracking
- An ongoing monitoring and escalation plan

LLM Listed does not claim to control the answers produced by independent AI platforms. AI responses are probabilistic and can change according to the platform, model, context, location, sources and date of testing.

Instead, the service documents material discrepancies and improves the authoritative information available to the systems producing those answers.

Harper said:
“This is not another AI visibility score.
“A visibility report tells a company how often it is mentioned. Brand Protection tests whether specific, verifiable statements about the company are accurate.

“The distinction matters. Being overlooked is a growth opportunity. Being confidently misrepresented is an immediate brand and commercial risk.”

A new responsibility for companies

Responsibility for AI business accuracy does not currently sit neatly within one corporate department.

Product marketing may own facts about features and customer suitability. Communications may be responsible for company narratives and leadership information. Legal, compliance or risk teams may become involved when AI makes potentially harmful or regulated claims.

LLM Listed believes companies will increasingly need a defined internal process for:

- Establishing which company facts are commercially material
- Monitoring how AI platforms represent those facts
- Recording and escalating repeatable inaccuracies
- Correcting conflicting information at source
- Retesting after material company changes
- Monitoring whether previously resolved claims return

Companies that have recently changed prices, launched products, completed acquisitions, entered new markets, secured certifications, appointed executives or rebranded may face an elevated risk of outdated answers.

“AI business accuracy is becoming a new form of brand governance,” Harper added.

“Companies already monitor what the press, search engines, review platforms and social media say about them. AI assistants now deserve the same scrutiny because they synthesize those sources into a direct answer that customers may treat as advice.”

About the research

LLM Listed evaluated 250 companies across 20 industries using ChatGPT, Gemini, Claude and Perplexity.

The sample included nationally recognized brands and randomly selected local businesses. Each platform was asked a consistent set of questions covering business identity, founders, headquarters, products and services, pricing, recommendations and reputation.

Responses were manually reviewed for:

- Factual accuracy
- Completeness
- Cross-platform consistency
- Recommendation behaviour
- Source attribution

A response could be broadly correct while also receiving an omission flag. For example, AI might correctly describe the primary purpose of a business but omit an important product or specialist service. Accuracy and completeness measures are therefore not necessarily mutually exclusive.

AI responses can vary between sessions and over time. The research represents the platforms’ answers during the stated testing period and should not be interpreted as a permanent assessment of any platform or company.

The full report is available at:

https://llmlisted.com/research/ai-business-accuracy-report-2...

Information about LLM Listed Brand Protection is available at:

https://llmlisted.com/services/brand-protection

About LLM Listed

LLM Listed helps organizations understand and improve how they are represented, cited and recommended across AI platforms including ChatGPT, Gemini, Claude and Perplexity.

Its work covers AI brand accuracy, AI visibility, Answer Engine Optimization, Generative Engine Optimization, AI reputation management and the information sources influencing AI-generated answers.

LLM Listed has delivered work for organizations including AMEX, ValiMail and Unbiased.

Media contact
Ben Harper
Founder, LLM Listed
Email: hello@llmlisted.com
Website: llmlisted.com

Available for interview on:

- AI misinformation about businesses
- How AI platforms influence company discovery and selection
- The commercial consequences of incorrect AI answers
- Why different AI assistants disagree about the same company
- How companies can monitor and correct AI representations
- The study’s industry, company-size and platform findings