CINC Develops 'AI Misinformation Check Function' to Verify Accuracy of ChatGPT, Gemini, and Perplexity Outputs

CINC Inc. (Headquarters: Minato-ku, Tokyo; President and Representative Director: Tomonori Ishimatsu; hereinafter 'CINC') has added an 'AI Misinformation Check Function' to its internal AI Search Optimization (GEO/LLMO) tool. This function enables users to verify whether generative AI responses contain incorrect information about their own company or products. With this function, users simply register the correct information about the company, service, or product they wish to verify. The tool then simultaneously sends queries (prompts) to ChatGPT, Gemini, Perplexity, AI Mode, and AI Overviews, and retrieves the responses. The acquired response data is automatically classified using a seven-level labeling system including 'Accurate', 'Context Mismatch', 'Inaccurate', and 'Hallucination', and is displayed alongside the cited URLs that served as information sources for the response generation. This function enables CINC to provide smooth support in its AI Search Optimization (GEO/LLMO) consulting services, from 'understanding the status of misinformation occurrence' to 'verifying the effectiveness of implemented improvement measures'. The daily use of generative AI such as ChatGPT and Gemini has brought to light the risk of AI providing incorrect information without companies' knowledge. In the World Economic Forum's 'Global Risks Report 2026 Edition', 'Misinformation and disinformation' is ranked as the second most severe risk to the world over the next two years. Translating the risk of misinformation into familiar corporate examples, cases include 'AI continues to recommend the old model even after a product renewal' or 'Outdated prices are included in responses after a price revision'. These phenomena occur because all articles on the web, including press releases, reviews, and comparison articles, are read by AI, forming a brand's 'image within AI'. The risk that unintentionally accumulated old information or errors are used in responses to users exists for any company. Main Business Risks Caused by AI Misrecognition Type of Risk / Specific Expected Events - Guidance to Discontinued/Old Models: AI continues to recommend old models after a product renewal, leading to inquiries about non-existent products and loss of sales opportunities. - Complaints Due to Old Prices/Specifications: AI responds with old information after price/specification changes, leading to complaints or cancellations from users who considered purchases or contracts based on incorrect premises. - Guidance to Old Addresses/Contact Information: AI continues to guide users to old locations after a relocation, causing trouble with visits and deliveries. - Hallucination-Based Misinformation: AI fabricates and responds with non-existent features, awards, or capital relationships, posing a risk that users accept this as fact. - Mismatch Due to Misrecognition of Recruitment Information: AI responds with salary ranges or internal systems that differ from reality, potentially causing candidates to drop out before or after applying. - Brand Context Mismatch: AI mentions the brand in a context different from its intended strengths or confuses it with competitors, leading to unintended opportunity loss. Risks like those above are often overlooked because companies do not know 'what AI is answering'. This function was developed to visualize such 'invisible risks' and consistently support clients from understanding their current situation to measuring the effects of implemented measures. The 'AI Misinformation Check Function' enables the following support: 1: Comprehensive Understanding of AI Misrecognition By registering the correct information about the company, product, or service to be verified, queries (prompts) are simultaneously sent to five services: ChatGPT, Gemini, Perplexity, AI Mode, and AI Overviews, and the respective responses are automatically collected. Users can get an overview of which AI is providing incorrect answers for which items. 2: Labeling the 'Type' of Misinformation and Designing Countermeasure Priorities Instead of simply judging responses as accurate or inaccurate, it is possible to identify them using a seven-level labeling system: Accurate, Mixed Freshness, Context Mismatch, Incomplete, Inaccurate, Hallucination, and No Information. The appropriate countermeasure differs depending on whether the response content is 'factually different', 'information is old', or 'context is off'. Identifying by these seven levels leads to improved precision of proposed measures. For example, when querying about the latest color variations of a product, the responses from ChatGPT, Gemini, Perplexity, AI Mode, and AI Overviews are labeled as 'Inaccurate', 'Mixed Freshness', or 'Accurate', making it immediately clear what deficiencies exist in each response. Additionally, by checking the cited links along with the response details, it is possible to identify the source of the incorrect information. This function can also acquire data when generative AI responds without searching, making it possible to investigate whether the cause of misinformation is the AI's internal knowledge or the cited web pages. Regular checks support early detection and handling of client risks. CINC's AI Search Optimization (GEO/LLMO) consulting provides end-to-end support from problem analysis and strategy formulation to measure execution and improvement, aiming for a state where generative AI cites the client's website or recommends the brand in an appropriate context. For those who 'first want to check their current situation', CINC offers a free diagnostic service for AI Search Optimization (GEO/LLMO). Applications can be made via the following URL: https://consulting.cinc-j.co.jp/geo_diagnosis ■ About CINC's AI Search Optimization (GEO/LLMO) https://www.cinc-j.co.jp/service/analytics/geoconsulting ■ CINC's Marketing DX Services Digital Marketing Strategy Design Consulting Content Marketing Consulting AI Search Optimization (GEO/LLMO) Consulting SEO Consulting SNS Consulting Conversion Improvement Consulting Advertising Operation Agency YouTube Marketing Consulting YouTube Channel Operation Consulting YouTuber Tie-up Strategy Design & Distribution ■ About Keywordmap Keywordmap is a tool equipped with functions to support SEO and content marketing, including competitive analysis, keyword selection, AI writing, and effect measurement, based on data from a vast number of search results, own/competitor/all websites, and listing ads. Since its release in 2016, it has been used by a wide range of companies across industries and business scales to drive web traffic. URL: https://keywordmap.jp/ Keywordmap Free Trial: https://keywordmap.cinc-