Last updated: 24.08.2026

TL;DR: AI search engines don't just prefer earned media over brand and social content. They also prefer brands you've already heard of. A recent University of Toronto study found that well-known and niche brands get meaningfully different treatment across ChatGPT, Claude, Gemini and Perplexity, with AI models agreeing far more often on established names than on challenger ones. If you're not a household name, being technically "visible" to AI isn't enough. You need a deliberate strategy to maneuver your way onto the shortlist.

I spend a lot of time playing around with AI systems. It’s a perk of the job. If you’re a niche or independent business, I’d bet cold hard cash that searching for your brand on AI tools will go something like this… I’d ask a question like "What's the best [your category] out there?" and the AI assistant would confidently rattle off three names. Two of them would be the brands that were already famous before AI search existed. The third would be whoever has spent the most on earned coverage this quarter. As the smaller player, even if you get better results, you wouldn’t appear in the answer at all.

That's not bad luck. It's measurable behaviour. But it can be overcome if you know how to game the system. 

The bias isn't just about media type; it's about brand size

Dip your toe into GEO and you’ll read about how AI search overwhelmingly favours earned media, independent reviews, comparison sites, and expert coverage over brand-owned and social content. A 2025 University of Toronto study analysing ChatGPT, Claude, Gemini and Perplexity confirmed this at scale. AI engines pulled the overwhelming majority of their citations from third-party sources rather than brand websites, often excluding social platforms like Reddit almost entirely.

What's less talked about is the second layer of that bias: AI engines don't treat every brand equally, even when the media mix looks similar on paper.

The same research pitted well-known brands (think Apple, Nike, Toyota-tier names) against niche or specialist brands in the same categories, then checked how consistently different AI engines agreed on the answer. For well-known brands, agreement between models sat between 76 and 81%. For niche brands, it dropped to 71–76%. A five-point gap doesn't sound dramatic until you realise what it means in practice. The more established you are, the more likely every AI tool your customer opens gives them the same answer, i.e., your name. The less established you are, the more your visibility depends on which specific engine, and which specific prompt, someone happens to use.

The source mix shifts too. ChatGPT leaned on earned sources for roughly 93.5% of citations about well-known brands, and pushed that even higher, to 95.1%, for niche ones, proving brand-owned content can’t stand alone. Claude followed a similar pattern. Perplexity was the most forgiving of the group, routing more of the share of niche-brand answers through social and brand sources rather than earned exclusively.

Translation: if you're not already famous, third-party validation isn't a nice-to-have. It's the main door in. Like those speak-easy bars where the door is disguised by a bookcase or fridge. When you open it up, it’s like stepping into another world. I’d love to go to one of those bars!

Quick wins we wish would work vs. what actually gets challenger brands cited

In AI search visibility, there isn’t really any ‘just do X, Y, Z’ advice I can confidently give because it’s never one and done. It’s an ongoing strategy. 

Five ways challenger brands can maneuver their way onto the AI-generated answer shortlist

The big brand and earned media bias exists. Here's what to actually do about it.

1. Prioritising earned coverage while building brand traffic. AI engines are researching niche brands on earned media so you need a PR and outreach strategy to make sure you have the backing of third-party sources. Credible mentions on a trade publication or comparison site are just as important as maintaining your own blog.

2. Get named inside comparison content, not just your own reviews page. AI models build shortlists by pulling from articles that already compare multiple options in your category. Being featured in someone else's "best of" roundup carries more weight than publishing your own, because it's a source the AI has already learned to treat as neutral. If decipher. ever publishes a blog on the best GEO agencies, you’ll know we’ve been hacked!

3. Don't rely on a single AI engine to "get" you. Agreement between models is lower for niche brands precisely because each engine sources differently. Perplexity pulls more from social and brand domains while ChatGPT and Claude lean more on earned. Test how your brand actually shows up across ChatGPT, Claude, Gemini and Perplexity individually rather than assuming one strong result means you're covered everywhere.

4. Make the case for you explicit, not implied. The study also found that AI answers increasingly function as justified shortlists. AI needs a reason to recommend you, not just proof you exist. Give it a juicy detail it can use in comparison tables against competitors, such as "longest warranty in category" or "best for small teams," and publish structured content about it so those claims are easy to lift.

5. Build a presence before you need it. Big brand bias compounds. The more an AI engine has already cited you, the more likely it treats you as a safe, recurring answer. The sooner you start building outside coverage, the better, especially as earned authority is not something you can switch on overnight.

Why this matters more for challenger brands than the headline stats suggest

• You don't get graded on a curve. Established brands can coast on residual recognition even with a thin earned-media footprint. Niche brands don't have that buffer. The study's numbers suggest AI engines default to caution around less familiar names, which in practice means defaulting to silence.
  
  

• The gap shows up exactly where AI search is growing fastest. Consideration-stage queries, "best X for Y," "X vs Y", are where challenger brands most need to appear, and where the research shows earned media dominates most heavily across every engine tested.
  
  

• Not everyone is doing this yet. Most independent and mid-sized brands are still optimising for Google rankings and treating AI visibility as an afterthought. The brands that start building deliberate earned-media strategy now will be the ones AI engines default to once the category gets crowded.
  
  

• It compounds against you if you ignore it. Every quarter that a competitor accumulates earned mentions while you don't is a quarter that gap gets harder to close, because AI engines are increasingly consistent about who they've already learned to trust.

Key takeaways

• AI search doesn't just favour earned media over brand and social content; it treats well-known and niche brands differently even within that pattern.

• Multi-engine agreement on well-known brands (76–81%) is meaningfully higher than on niche brands (71–76%), meaning your visibility is less predictable the less established you are.

• ChatGPT and Claude route niche-brand answers through earned sources at rates above 85–95%, all but excluding brand-owned content.

• Perplexity is comparatively more forgiving of smaller brands, which means engine-by-engine testing matters more than ever.

• Closing the gap is a content and outreach problem, not a technology one, but it takes sustained earned-media effort, not a single campaign.

Stop waiting to be recognised. Start being cited.

Big brand bias isn't a reason to give up on AI visibility if you're not a household name yet. It's a reason to be more strategic about your visibility. The brands that win here aren’t only being consistent on their own channels. They’re the ones showing up, consistently and credibly, on the sites AI engines already trust.

At decipher., we help challenger and independent brands build the kind of earned-media footprint and structured content that gets picked up across ChatGPT, Claude, Gemini and Perplexity, not just the one engine that happens to like you. If you want to stop losing shortlist spots to bigger names with thinner cases, let's talk.