AI-driven discovery isn't neutral. A new study shows models search for brands they already know 3.2 times more often than unfamiliar ones, and the implications for challenger brands are serious. Here's a number worth considering: 55.7% of the time an AI model ran a brand-related search, it searched for a brand it already knew. Brands outside the model's top 10 most familiar accounted for only 17.4% of searches, a 3.2× gap, according to a GeoSurge study analyzing 3,960 AI responses across 66 U.S. buyer prompts. If you're running demand generation for a mid-market SaaS company that isn't Salesforce or HubSpot, that ratio should concern you.

The Winner-Take-More Loop

Understanding the study's methodology is crucial. Researchers measured what the model already "knew" about a brand (its memory) and what it chose to search for when answering buyer questions. The correlation was stark: 63% of brand-specific searches involved one of the model's five most familiar brands. This concentration suggests a feedback loop that is difficult to break from the outside. Models train on historical data, which over-represents established brands due to their greater coverage and documentation. Consequently, the model searches for these brands when answering queries, generating more engagement data that feeds back into future training. Incumbents become more visible, while challengers remain marginalized. Practitioner commentary on this study is blunt: AI agents "do not start from a neutral slate." Brand familiarity skews discovery toward companies with existing digital prominence.

Industry Variance Is Real (and Worth Checking)

One detail often overlooked is that the familiar-brand bias ranged from 41% to 82% depending on the industry, with unfamiliar-brand searches ranging from 9% to 23%. This wide spread includes business software, travel, automotive, finance, luxury, fitness, food, education, and fashion. Before panicking or dismissing this, validate within your own vertical. A B2B SaaS company targeting mid-market RevOps teams faces a different competitive set than a luxury brand. While the bias exists in every measured sector, its intensity varies enough that blanket assumptions won't aid your planning. It's also important to acknowledge that familiarity bias may reflect real-world trust signals. Well-known brands often have more third-party citations and coverage. Models might be reflecting available evidence rather than arbitrarily favoring certain brands. This doesn't change the competitive impact on challengers but clarifies where to focus efforts: building a digital presence that provides models with cleaner, richer signals about your brand.

What to Actually Do About It

The study revealed one encouraging data point: in at least one instance, Gemini searched for Lemon Squeezy while answering a question about online payment providers, even though that brand wasn't in its measured memory. Live search can still surface unfamiliar brands, especially in categories where models rely less on stored knowledge. The door isn't closed; it's just narrower than many teams assume. Practitioner guidance recommends several concrete steps. First, audit how AI models describe your category and competitors. Run prompts across ChatGPT, Gemini, and Claude to see who gets mentioned, how they're described, and where your brand is absent. Second, strengthen entity clarity: ensure your brand's key associations (what you do, for whom, what makes you different) are consistent across your site, earned media, and third-party sources. Third, keep content current; models trained on stale data will reflect outdated brand associations. For Marketing Ops teams, consider adding a new metric: "AI Share of Voice." This measures how often your brand appears in AI-generated answers for category and problem queries. This signal won't show up in traditional SERP rankings or paid dashboards; you need to actively seek it out.

The Measurement Gap

This is the part most teams will overlook, yet it matters most. Traditional attribution frameworks fail to capture AI-mediated discovery. If a buyer asks ChatGPT, "what's the best pipeline attribution tool for mid-market SaaS" and your competitor is named multiple times while you are not mentioned, that's a top-of-funnel visibility loss that won't appear in your CRM. No UTM parameter, no referral source—just a buyer who never considered you. Ignoring this trade-off means your paid and content programs might perform well on their own metrics while your consideration-stage pipeline quietly shrinks. This issue can take two quarters to diagnose and another two to fix. The GeoSurge researchers noted that their results show a relationship between memory and search behavior, not proven causation. Fair enough. But a 3.2× gap in search frequency between familiar and unfamiliar brands across nearly 4,000 responses is a strong signal to act on. Brands that treat AI visibility as an operational metric now will gain a compounding advantage over those that wait for more definitive data. By then, the feedback loop will have already done its work.