One of the largest U.S. fitness brands ran an experiment earlier this year. They queried their key category terms across AI platforms to test visibility. The result: a small local gym in Houston was outranking them in AI search. Not occasionally. Consistently. This is from a piece by researchers at MIT and the University of Virginia’s Darden School, published in MIT Sloan Management Review in January 2026.
The fitness brand had the bigger marketing budget. It had the better-known name. It had invested heavily in traditional SEO for years. None of that mattered when ChatGPT decided who to recommend.
This is the story playing out across B2B right now, and most teams discover it the wrong way: during a sales call, when a prospect mentions they already checked with ChatGPT and got three vendor names — and yours wasn’t one of them.
Before you commission a GEO audit of your owned content, there are three channels to check first. Skip these and you’re optimizing the wrong thing.
AI engines don’t pull citations from websites the way Google pulls rankings. According to Moz’s analysis of 40,000 queries in 2026, 88% of Google AI Mode citations come from URLs that are not in the organic top 10 results. The traditional SEO signal — your domain authority, your ranking position — has almost no correlation with whether you appear in AI-generated answers.
What does correlate: brand mentions from third-party sources. Ahrefs’ analysis of ChatGPT’s citation behavior found that branded web mentions correlate at 0.664 with AI overview visibility, compared to 0.218 for backlinks. That’s a 3x difference. The signal that drives AI citations is not your website — it’s whether other publications and platforms are talking about you.
This is why the Houston gym beats the national fitness chain in AI search. It’s been mentioned in local press, reviewed on high-traffic platforms, cited in community forums. The chain has better SEO. The gym has more citations.
The three-channel audit below is designed to surface exactly this gap before you build a GEO plan on top of a broken foundation.
Generic category queries (“what is [category]?”) are not how your buyers find vendors. They ask comparative, specific questions: “best [your category] for [your ICP]”, “top [solution type] for [industry] companies under [size]”, “[your product type] vs [alternative]”. These are the prompts where vendor citations actually happen.
Pull 15-20 of these from your sales team. The questions your reps hear before deals go dark are usually the same questions buyers are running through ChatGPT before they make a list.
Run each query through ChatGPT, Perplexity, and Google AI Overviews. For each response, log: which brands appear, how high in the response, and what source they’re cited from. Do this in a simple spreadsheet — you don’t need a tool for the initial pass.
What you’re building is a citation map: who’s in the answers, and more importantly, what’s getting them there. The source is the signal.
Informa TechTarget’s March 2026 launch of their AI Visibility Audit product cited that B2B buyers are completing vendor shortlists inside AI platforms before contacting any company directly — with up to 60% of searches ending without a click through to any website. By the time your SDR reaches out, the buyer has already run these queries. Your placement in the answers — or absence from them — has already shaped their perception.
Give this channel 30-45 minutes. That’s enough to know whether you have a problem.
For every competitor that surfaces in the queries above, identify the specific sources pulling them into the AI answers. Not the competitors’ own websites — the third-party pages, reviews, and publications that are being cited.
You’re looking for patterns. Does a competitor appear because they’re consistently covered in specific trade publications? Because they have a strong G2 or Capterra presence? Because they’ve been cited in industry analyst reports? Because journalists at certain outlets keep referencing them?
Muck Rack’s generative AI citation analysis found that over 85% of non-paid AI citations originate from earned media — third-party coverage in publications AI engines already treat as authoritative. Press releases, product pages, and owned blog content account for a tiny fraction of what gets cited.
This means your competitor’s AI citation advantage is not a content strategy advantage. It’s a coverage advantage. They’ve accumulated more mentions in the right places.
Cross-reference this against the sources currently citing your brand. The gap between their citation footprint and yours is the actual work to be done.
The GEO-16 framework (Kumar et al., arXiv, September 2025), which audits 16 on-page signals across 70 B2B intent prompts, found that even pages scoring high on technical quality signals had materially lower citation rates when they were on vendor-owned domains rather than third-party publications. The findings documented that AI search systems “systematically favour earned media (third-party, authoritative domains) over brand-owned and social content.” On-page optimization matters — but it’s downstream of whether you’re in publications the AI engine trusts.
AI engines show strong citation bias toward content from high-authority domains. Ahrefs’ analysis found that 65.3% of ChatGPT’s most-cited pages came from domains with a DR of 80 or above. This is the threshold that matters.
The question to answer for your brand: how many DR 80+ domain mentions have you accumulated in the last 12-18 months, and in which publications?
You don’t need a tool for the initial pass. Run your domain through Ahrefs or a comparable backlink tool, filter for referring domains above DR 80, and look at whether those domains are editorial publications — Forbes, TechCrunch, MIT Sloan, industry trade press with long track records — or just directory listings and aggregators.
Directory listings don’t drive AI citations at the same rate as editorial coverage. The Muck Rack data is clear: earned media from the publication types AI engines were trained on — journalism, research institutions, authoritative industry outlets — is what moves the needle.
If the bulk of your DR 80+ mentions come from directories, aggregators, or press release syndication rather than genuine editorial coverage, you have a citations gap that no amount of on-page GEO optimization will close. You can build the cleanest FAQ schema and the most structured heading hierarchy in your category and still not appear in AI answers, because the domain signal isn’t there.
The MIT Sloan researchers make this point directly: even brands with the largest market shares and the highest traditional search investments can become invisible in AI-driven search if they haven’t built the right citation infrastructure. In the financial services case they document, the market leader had invested the most in digital marketing, had the highest brand recognition, and still wasn’t recommended — a smaller competitor with a more distributed citation footprint was.
The most common mistake right now is running GEO optimization as if the primary problem is on-site. Teams audit their website structure, add FAQ schema, restructure headings for extractability — and then wonder why nothing changes in AI answers.
All of that work is necessary. The Princeton/Georgia Tech GEO research (Aggarwal et al., SIGKDD 2024) found that adding statistics to content improves AI visibility by 30-40%. Structure matters. Schema matters. Freshness matters. Gartner projected a 25% decline in traditional search volume by 2026 as AI tools absorb that traffic.
But on-page optimization compounds only when you have citation infrastructure underneath it. The brands building citation authority now are not just doing GEO on their own sites. They’re getting editorial coverage in publications AI engines have been citing for years.
Run the three-channel audit before you build the GEO roadmap. The roadmap will look different once you know where the gap actually is.
The earned media signal driving AI citation rates is not a coincidence. AI engines learned to trust the same publications that shaped editorial credibility with human readers for decades — Forbes, TechCrunch, MIT Sloan, industry trade press with long track records. These are authoritative because they were authoritative before AI systems existed.
This is what Machine Relations describes as the foundational layer of brand authority in the AI era: earned media in trusted third-party publications is not a GEO tactic — it’s the infrastructure the tactics sit on top of. The mechanism that made PR valuable in the first place is now the mechanism determining which brands AI recommends. On-page signals compound on top of it. Without it, you’re optimizing air.
The three-channel audit tells you where your foundation stands. Run it before the next GEO budget conversation.
If you want to see exactly how your brand currently shows up across AI engines — before you decide what to fix — the AuthorityTech visibility audit gives you that baseline in one pass.