Google's Gemini Is Making Up Things About Your Business. There's No Fix Yet.

Somewhere right now, a potential customer is asking Gemini about your company. And there's a decent chance the answer includes something that never happened. A lawsuit you weren't part of. A product recall that was actually your competitor's. A regulatory action that somebody at Google's model decided sounded plausible enough to state as fact.

An ecommerce business owner posted on Reddit this weekend describing exactly this. Gemini had been generating false claims about their company, including fabricated legal actions. The post blew up, 66 upvotes and 45 comments, mostly from other business owners sharing similar experiences. They aren't isolated cases. Wolf River Electric, a solar contractor, says Gemini repeatedly told users the company was sued by a Minnesota attorney general for deceptive sales practices. That lawsuit never existed. The company claims client cancellations followed directly.

According to AI Labs Audit's 2026 research, 35% of brands now report that inaccurate AI responses have damaged their reputation. And most found out the way you'd expect: not from a dashboard, but from a confused customer.

This isn't a bug that gets patched

I think a lot of teams are assuming this problem will go away as the models get better. And technically, Gemini 3 did reduce hallucination rates on standard benchmarks. But benchmarks test general knowledge. They don't test whether your specific business has a clean legal record, or whether you were the company involved in that product recall from 2023, or whether you actually operate in all the states Gemini says you do.

The model doesn't know these things. So it guesses. And sometimes the guess is confident, specific, and completely wrong.

In one client's case we looked into recently (mid-size DTC brand, roughly $4M in annual revenue), Gemini was attributing a competitor's product recall to them. Similar company name, similar product category, and the model just... merged the two entities. The client's customer service team started getting calls about it before anyone on the marketing side knew it was happening.

The frustrating part: there's no correction mechanism

With Google Search, you can file a removal request. You can update your site, disavow bad links, manage your Knowledge Panel. There are established processes, imperfect as they are.

With an LLM, there's nothing equivalent. You can't submit a correction to Gemini. There's no "dispute this claim" button. The model's knowledge comes from its training data and whatever retrieval augmentation Google layers on top, and you don't have direct access to either channel.

The options right now are limited and, honestly, a bit unsatisfying:

Strengthen your structured data across every surface you control. Schema markup, Google Business Profile, Wikipedia (if you have a page), Wikidata, Crunchbase. The more authoritative, structured sources that contain accurate information about your business, the more likely the model is to surface correct facts. This part is worth doing regardless.

Monitor what AI chatbots say about you. Tools like AI Labs Audit and Originality.ai now offer brand monitoring across major LLMs. It's a new category of tool that didn't exist a year ago, and I expect it to grow fast. Setting up basic monitoring takes about 20 minutes. At minimum, run a manual check monthly: search your brand name, your founder's name, and your top product names across Gemini, ChatGPT, and Perplexity. Keep screenshots. You'd be surprised how often the results change between queries, which makes the documentation part even more important.

Document everything. Lawsuits against AI companies for defamation are stacking up. Analysts expect at least two more major suits this year. If your business is being harmed by false AI-generated claims, having a documented record of what was said and when matters for any legal action down the road.

Meanwhile, Gemini's influence is growing fast

Here's the part that makes this more urgent than it was six months ago. SE Ranking's January 2026 data shows Gemini now sends 29% more traffic to websites than Perplexity. In the US specifically, the gap is 41%.

That's a complete reversal. Five months earlier, Perplexity was sending nearly 3x more traffic than Gemini. The 115% surge coincided with the Gemini 3 deployment in December, and it hasn't slowed down. ChatGPT still accounts for roughly 80% of all AI referral traffic, but Gemini is the one gaining ground fastest.

Which means more people are seeing whatever Gemini says about your business. More people forming opinions based on answers that might be fabricated. The reach of the problem is expanding even if the per-query error rate goes down.

Also worth watching: CTV's measurement problem and Walmart's Shopping trick

Two other stories from this week that practitioners should have on their radar.

CTV advertisers are paying for impressions that run after viewers turn their televisions off. The streaming device stays on, keeps serving ads, and measurement systems don't flag it. Global CTV ad fraud losses are estimated at $7.4 billion. OpenX and TVision moved attention data into pre-bid filtering in March, which helps if your DSP supports it. Most buyers haven't activated it yet. If you're spending on CTV, ask your platform partner about pre-bid attention signals this week.

And in an interesting piece of competitive intelligence, Walmart operates over 1,600 distinct seller names in Google's organic Shopping carousels, each one linking back to Walmart.com but counted by Google's diversity algorithm as a separate retailer. A Productrise study analyzing over a million product listings found that grouping all Walmart-affiliated names as one entity drops "true diversity" in Shopping carousels by 32%. If you sell on Google Shopping and noticed your organic visibility declining, this is probably part of the picture. Not much you can do about Walmart's strategy, but understanding why your impressions shifted helps you decide where to reallocate effort.

Search your brand name in Gemini before someone else does

If there's one thing to do today, it's this: open Gemini, ChatGPT, and Perplexity. Ask each one: "Has [your company] been involved in any lawsuits?" and "What are common complaints about [your company]?" See what comes back.

In a few cases you'll find the answers are fine. But in enough cases to be alarming, you'll find confidently stated facts that are partially or entirely wrong. And the sooner you know about it, the sooner you can start building the structured data foundation and monitoring infrastructure to manage it. Because this isn't a problem that's going away. The models are getting more popular, not less. And right now, nobody is checking whether what they say about your business is true.