AI Knows Nothing About You. It Describes Someone Else.

What AI does when it doesn't know you

Ask ChatGPT or Gemini about your company. If you're a major brand, the answer will be reasonably accurate. If you're not, the answer will still sound reasonable. It just won't be about you.

A recent piece on Search Engine Journal highlights a problem we see increasingly often with our clients: when AI models lack sufficient information about a company, they don't decline to answer. They substitute. With competitor data, outdated information, or industry-level truths incorrectly pinned to your name.

The most troubling part isn't the substitution itself. It's that you can't discover it through a standard content audit. The issue doesn't live in what you've published. It lives in what's missing about you, in places you don't control: press mentions, business directories, reviews, third-party references. That's where AI models look when building answers about brands.

Four substitution patterns you should know

Duane Forrester, founder of UnboundAnswers and former Bing executive, identifies four patterns through which AI fills information gaps. We see them regularly when testing brand visibility in AI at difrnt., and each one requires a different response strategy.

Silent analogy. The model describes a competitor as if it were your company. It doesn't explicitly name the competitor but borrows their characteristics and attributes them to you. If you're not actively checking what AI says about your brand, you'll never notice. The answer sounds plausible. It's just not yours.

Staleness presented as current. AI uses two-year-old data as if it were fresh. If you've changed your positioning, services, or team, the model doesn't reflect those changes. And it presents everything with the same confidence it would use for last week's data. A potential client searching for you in AI gets a picture that no longer matches what you actually do.

Thin evidence delivered with maximum conviction. A single source becomes "consensus." A blog post written by a third party three years ago becomes the foundation for your company's entire AI description. No nuance, no context, no update. The model can't distinguish between an isolated source and a verified fact confirmed from multiple directions.

Category knowledge applied to your name. The model knows how your industry works in general. It knows what a marketing agency does, what a consulting firm does, what a SaaS platform does. But it doesn't know what makes you different. So it describes the industry and puts your name on it. The result looks correct at first glance. It just doesn't represent you, and it doesn't differentiate you from any competitor.

Why publishing more content won't fix this

The instinctive response to this problem is predictable: "let's publish more." It's every marketer's natural reflex. More articles, more pages, more presence. But academic research cited by Forrester, including studies from EMNLP 2021 (Mallen et al., Longpre et al., Sciavolino et al.), shows why it doesn't work that simply.

AI models have a popularity bias built into their architecture. Entities with abundant data from varied sources get described accurately. Those with limited data get described by analogy with popular ones. And the retrieval mechanism (RAG) doesn't correct this bias. It applies it a second time, because it selects from the same highly visible sources.

In practice, if you're a small or mid-sized brand, publishing 10 or 20 more blog posts won't fundamentally change how AI describes you. You need to change where and how you appear in the sources AI actually consults. You can't control the model, but you can influence the data it learns from.

How to check if AI is substituting your brand

You can't audit this from Google Analytics or Search Console. It doesn't appear in any standard dashboard. You need to do something few marketers practice systematically: interrogate AI models directly about your brand and compare the answers to reality.

Here's what we recommend at difrnt.:

Ask category questions, not brand questions. Instead of "what does company X do?", ask "who are the best digital marketing agencies in Romania?" or "who provides SEO services for e-commerce?". Check if you appear. And if you do, verify whether the description is accurate or whether the model has substituted information.

Test with comparative questions. "What's the difference between company X and company Y?" This is where substitution becomes obvious: if AI can't differentiate the two companies, it means it's pulling the same generic information for both. That's a clear signal your brand lacks sufficient digital identity in the sources models consult.

Repeat periodically. AI answers aren't stable. What appears today may vanish tomorrow, because models periodically update the sources they draw from. A single check isn't enough. We recommend monthly checks with a standard set of 5-10 questions.

Treat results as a gap map. Every incorrect or absent answer is a signal about where descriptions of your company are missing from third-party sources. Not from your own content (you're probably fine there), but from press coverage, business directories, reviews, interviews, partner case studies, and relevant mentions across the web.

What you can do right now

Our approach at difrnt. for clients facing this problem combines GEO (Generative Engine Optimization) with digital PR and earned media. It's not a one-month project, but every step counts:

Build presence in third-party sources. AI doesn't only read your website. It reads everything written about you across the web. If nobody writes about you independently, AI reads your content but doesn't cite you. Or worse, it confuses you with someone else. Look for earned media opportunities: press articles, guest posts on industry publications, partnerships with industry organizations, conference appearances documented online.

Be specific in everything you publish. Explicitly named entities (people's names, products, locations, concrete numbers, years, clients) are harder to substitute than generic statements. "Digital marketing agency" could be anyone. "Agency founded in 2018 in Bucharest, specializing in SEO and PPC for e-commerce, with clients in retail and fintech" is much harder to confuse with someone else.

Monitor AI visibility monthly. Add a dedicated section to your monthly reporting: how do ChatGPT, Gemini, Claude, and Perplexity respond to 5-10 relevant questions about your brand? It's the only way to detect substitution before a prospect discovers it instead. And if you don't have time to do this internally, find a partner that offers GEO monitoring.

FAQ

What does AI substitution mean for a brand?

Substitution happens when an AI model lacks sufficient data about your company and fills the answer with information from a competitor, outdated data, or generic industry truths. It doesn't refuse to answer. It delivers a response that looks correct but doesn't represent you.

How do I find out if AI describes my company accurately?

Directly query ChatGPT, Gemini, and Perplexity with category and comparative questions. Compare the answers with what you actually do. Repeat monthly, because answers change over time.

Does publishing more blog content fix the problem?

Not entirely. AI models have a popularity bias that favors entities with many mentions from diverse sources. Your own content helps, but presence in third-party sources (press, directories, reviews) matters at least as much in preventing substitution.