I was sitting with a client last month, looking at why ChatGPT kept recommending their competitors. Their website was polished. Their SEO was solid. They ranked well on Google. And when I asked ChatGPT for the best option in their category, it named three other companies and skipped them entirely.

So I dug into what the AI was actually reading when it built that answer. What I found stopped me: the AI wasn't reading their website at all. It was reading what other people said about them — on Reddit, YouTube, review sites, and industry directories. The conversation happening around them mattered more than anything they'd published themselves.

This isn't one engine being weird. An analysis of 107 million AI answers across ChatGPT, Perplexity, Gemini, and Google AI Overviews found that community and user-generated content makes up roughly half of the leading citations in every industry studied. Reddit threads. YouTube reviews. Quora answers. Social posts. What your customers write about you — not what you write about yourself.

For B2B companies, it's starker. In software and IT services, brand-owned websites account for 2% of AI citations. Two percent. Independent editorial and reference sources — B2B directories, industry publications, third-party reviews — account for 47%. One separate analysis by Surfer's AI Tracker suggests YouTube alone may account for nearly a quarter of all AI citations across industries, with Wikipedia and Google.com also consistently dominant. Video appears to be the single most cited content format across every vertical.

The engine deciding whether you get recommended isn't reading your pricing page. It's reading a Reddit thread from two years ago where someone asked about your category and three people mentioned your competitor. It's watching a YouTube review. It's pulling from a directory listing. Your own site — the thing you've spent years optimizing — is a rounding error.

The old playbook was to build your website, optimize it, and wait. The new playbook is to make sure the conversation happening elsewhere — the one you don't control — is saying the right things about you. Because that's what the engines are actually reading.

What to do about it

Start with the platforms where the conversation happens. Reddit, YouTube, Quora — this is where AI goes to learn what real people think. If you're a B2B company and nobody's mentioned you on Reddit, or there's no YouTube content about your product, the AI has nothing to cite. Get your customers talking there, or create useful content yourself. Not ads — answers, demonstrations, reviews.

For B2B specifically, get onto the directories the engines cite. G2, Capterra, TrustRadius, industry-specific review sites — these independent editorial and reference sources make up 47% of B2B citations. If you're not listed, or your listing has three reviews while your competitor has three hundred, you're invisible. Claim every profile. Get recent customers to leave specific, detailed reviews — not "great product" but "we used it for our data pipeline migration and it cut processing time in half." Specifics give the AI something to cite.

Produce video. I know that's the annoying one if you're not a video person. But according to the data, video is the most cited content format across every vertical. YouTube tutorials, walkthroughs, comparisons, customer testimonials — it gives the engines something rich to pull from. It doesn't need to be polished. It needs to exist and be useful.

Finally, understand that the engines are not the same. What ChatGPT cites is not what Perplexity cites, and Google AI Overviews pulls differently again. Run the same query across all four and see what each one cites — that'll tell you exactly where your gaps are.