Gemini Uses Negative Words About 72.6% of Brands vs 57.2% for ChatGPT

A customer is about to sign with a moving company and asks ChatGPT a simple question: is this company any good? The answer describes the services, mentions a few strengths, and then brings up complaints. For most brands, that is what the answer looks like.

"Is [brand] good?" is one of the most common questions people ask before they buy, and AI assistants now answer it in a single paragraph. Whatever that paragraph says becomes the default description of a brand, and a mention of complaints or lawsuits in it reaches every customer who asks.

Gemini uses negative words for 72.6% of brands and ChatGPT for 57.2%. For about a third of brands, ChatGPT brings up negatives every time it is asked, and brands in auto sales and home services draw them most often.

In September 2026 we took 500 brands, the companies whose websites rank most often across a large set of Google searches, and asked ChatGPT and Gemini "Is [brand] good?" twice each, 2,000 answers in all. The brands span industries from auto dealers and movers to accounting firms and universities.

We checked each answer for terms from a fixed list of negative words, including complaints, scam, lawsuit, poor, fraud, recall, hidden fees and data breach. An answer counts as negative when it contains at least one of them. A brand counts as getting negative words when at least one of its two answers from an engine does, and "both runs" means both answers did.

Per answer, 44.8% of ChatGPT responses and 57.4% of Gemini responses contain at least one negative term. Because each brand was asked about twice, the brand-level rate is higher: 57.2% of brands for ChatGPT and 72.6% for Gemini. Gemini is the harsher of the two, bringing up a negative for nearly three brands in four.

For 32.4% of brands, ChatGPT used negative words in both answers. For those brands the negatives are not a one-off turn of phrase: they are part of how ChatGPT describes the company whenever someone asks.

The most common negative term by far is "complaints", used 723 times in ChatGPT's answers. Next come the singular "complaint" (75), "BBB complaints" (36) and "recalls" (27). ChatGPT's view of whether a brand is good leans heavily on what unhappy customers have posted, from review sites to Better Business Bureau records.

That tells a brand where the negatives come from. Complaint pages and review threads are among the sources AI engines read when judging a company, and an unresolved complaint can surface in an answer years after it was filed.

By industry, auto dealer brands draw negative words in 72% of ChatGPT's answers. Plumbing (66.7%), moving companies (65%), solar (60.3%) and HVAC (58.8%) follow. These are industries where customers make large, infrequent purchases, often in a hurry, and write detailed complaints when a job or a deal goes wrong.

At the other end, accounting firms draw negative words in 22.2% of ChatGPT's answers and higher education in 25%, the lowest of the industries in the table. Professional services with licensed practitioners and long client relationships get far gentler descriptions than trades where one bad job becomes a public complaint.

Ask ChatGPT and Gemini whether your brand is good and read what they say. If complaints or Better Business Bureau records come up, those are the pages the AI engines are reading. Respond to complaints publicly, resolve open BBB cases and ask satisfied customers to review you on the same sites, so the material AI engines draw on is balanced.

Then publish clear, factual pages about your pricing, policies, warranties and how you handle problems. Complaints and recalls are the most common negative themes, and a direct answer on your own site gives AI engines a source to quote alongside them. If you are in auto sales or home services, check your AI answers monthly. Our reputation management services include monitoring how AI engines describe your brand.

Embed this research

Cite this study