AI Is Describing Your Company Behind Your Back — Is It Being Honest?

The Second Reputation

A buyer types "best project management software for remote teams" into ChatGPT. Your company makes exactly that. But what the buyer reads has nothing to do with your website copy, your positioning doc, or the messaging your team spent three months refining. It comes from a G2 review posted two years ago, a Reddit thread you never knew existed, and a TechCrunch blurb that got your pricing wrong. That summary is their first impression of you. And you had no say in it.

Jaxon Parrott framed this precisely in a May 22, 2026 piece in Entrepreneur: AI engines are now explaining companies to buyers through category-level searches, generating what Parrott calls a "second reputation" — a parallel narrative assembled from third-party sources that operates entirely independently of anything you own or publish. Your website does not anchor it. Your press releases do not correct it. It exists in a layer of the buying process you cannot directly access.

The data behind this makes the stakes concrete. A 2025 Pew Research analysis of 68,879 Google searches found that 18% produced AI summaries — and when those summaries appeared, users clicked traditional results only 8% of the time. That second reputation is not supplemental. For a growing share of buyers, it is the only reputation that gets read.

Where AI Gets It Wrong

The problem is not that AI occasionally gets things wrong. The problem is the specific category of wrong it produces — confident, plausible, and sourced from real content that belongs to someone else.

In September 2025, Google AI Overview was documented serving defamatory content to users by misattributing competitor reviews to the wrong businesses. A buyer searching a service category could read a negative review about your company that was never actually written about your company. The review existed. The source was real. The attribution was not. That distinction does not survive the experience of reading it.

The misattribution problem has a structural cause. A May 2026 Muck Rack study analyzing more than 25 million links across ChatGPT, Claude, and Gemini found that earned media drives 84% of AI citations. The AI is not reading your about page and drawing conclusions. It is assembling your public identity from press coverage, third-party directories, forum discussions, and scraped review platforms — content written by people who may have been describing a different company, a different time period, or a different version of what you do.

The veteran ownership case from early 2025 illustrates how far this can drift. Google AI was observed rewriting business descriptions on listings, inferring ownership details like veteran status from unrelated posts — details that were never stated, never verified, just inferred and published as fact. The business never made that claim. The AI made it for them.

The Brand Name Test Fallacy

Most marketers who want to know how AI is describing their company do the same thing: they open ChatGPT, type in their brand name, and read whatever comes back. That test feels thorough. It is not.

Buyers do not search your company name. They search their problem. "Best social media scheduling tool for agencies." "Project management software for construction teams." "AI writing tool that sounds like me." By the time a buyer types your brand name into anything, they have already formed an opinion — often from exactly the kind of category-level summary described above. The brand name search tells you what AI knows about you after a buyer already cares. It tells you nothing about what AI is saying during the part of the process that actually shapes whether buyers care at all.

The Pew Research numbers make this concrete. Eighteen percent of Google searches produce AI summaries. When those summaries appear, users click through to traditional results only 8% of the time. That means the summary is not a preview of your content — for most users, it is a replacement for your content. The description is the impression. If AI is describing your category and your company appears in that description with wrong pricing, an outdated use case, or a misattributed review, there is no second chance built into that interaction.

Searching your own name in ChatGPT confirms you exist. It does not tell you what buyers are reading.

What You Can Actually Influence

if earned media drives 84% of AI citations, then earned media strategy is no longer just a visibility play. It is a direct input to whatever AI says about you in a category search six months from now.

The practical starting point is not your own brand name — previous sections already established why that test fails. Start one level up. Search your category the way a buyer would. "Best tools for agency social media scheduling." "Project management software for distributed teams." Read what AI produces before your company name appears at all. Identify which third-party sources keep showing up in those summaries — which publications, which review platforms, which forum threads AI appears to be drawing from. Those sources are the terrain. Your earned media strategy should be targeting them directly.

The Muck Rack finding is specific enough to be operational: 25 million links across ChatGPT, Claude, and Gemini, and 84% of citations traced back to earned coverage. That is where the input is. Which means a placement in a trade publication that covers your category is now doing two jobs — building visibility with readers and feeding the training diet that shapes what AI says when a buyer never clicks anything at all.

Audit what AI says about your category. Map which sources it cites. Treat those sources as the distribution channels they have become.