Is AI lying about your company? Here’s how to take control of AI-based results – 3 specific ways

Have you ever searched for your own company in ChatGPT or Perplexity? If not, do it now. You’d be surprised, but in a significant number of cases, the answer is inaccurate: it lists bad services, gives an old address, or worse, recommends your competitor instead.

Although website traffic from AI technologies (e.g. ChatGPT, Perplexity, Google Gemini) is currently only a fraction of traditional Google searches, the rate of growth is explosive (527% jump in one year). But the biggest concern is not future traffic, but current perception. If AI “understands” your company poorly, it’s like your reception is misleading visitors.

“Most brand descriptions generated by AI are still wrong.” – says Mihály Tóth, founder of KERKOM.

The solution is GEO (Generative Engine Optimization), i.e. optimization for generative engines. It doesn’t require a huge budget, just structured thinking. Here are 3 experiments that any SME manager can implement in 60-90 days so that machines can understand who you are.

1. Experiment: build an “LLM-friendly” topic cluster

Marketers have been building topic clusters (groups of articles organized around a main topic) for a long time, but AI requires us to change the structure. AI doesn’t read, it “scans” and mines data.

The task: Choose a topic that your company is strong in and transform the content so that it makes sense to a machine.

1. Using natural questions: Forget about keyword stuffing. The subheadings (H2) of your posts should be specific questions, as your customers would ask them in real life.
   • Bad: “Dental implant prices”
   • Good: “How much will a dental implant cost in 2025?”
2. Summary at the beginning: The first 100-150 words of your article should be a concise, to-the-point answer. There is no need for a long introduction (“In today’s fast-paced world…”). This is the section that AI uses most often as a resource.
3. Q&A Format: Use question-answer blocks.
   • Question
   • Short answer (1-2 sentences)
   • Details (2-3 paragraphs)
4. Technical background (Schema): Use the appropriate Schema markup (FAQPage, HowTo, LocalBusiness) to indicate what you are talking about at the code level.

What to measure? After 60 days, see if ChatGPT or Gemini are more accurately referring to you for targeted questions than before.

Experiment 2: Brand entity and “sentiment” audit

AI hates uncertainty. If it finds different information about your company (brand) on Facebook, in the business directory and on your website, it will “make up” something or mix it up with something else. We call this hallucination.

“If your brand information is scattered, the AI ​​will ‘Frankenstein-style’ piece together the answer from the first thing it finds.” – says Mihály Tóth, founder of KERKOM.

The task: Make your digital footprint consistent.

1. Record the “before” state: Ask the main models (ChatGPT, Perplexity, Gemini): “Who is [Company Name] and what do they do?”. Save your answers with a screenshot.
2. Site Cleanup: Make it crystal clear on your website’s About Us page what you do, where, and for whom.
3. Unify External Sources: Make descriptions consistent across Google Business Profile, Facebook, industry hubs, and partner pages.
4. Review Depth: The AI ​​doesn’t just look at stars, it also analyzes text reviews to determine sentiment. Ask satisfied customers to write in detail about the service.

A specific example: An air conditioning company that did 60% of its industrial cooling business was identified by the AI ​​as a “residential service provider.” After rewriting the Google Business Profile and website to focus on “industrial and residential,” the AI ​​learned the correct definition in 70 days.