Okay, I need to tell you about something that made me reorganise my entire content strategy at 1:47 AM on a Tuesday. And no, it wasn't another LinkedIn guru telling me to "build in public" (though I did see three of those before breakfast). It was a dataset. 238,000 citations. From ChatGPT. And the results are so counterintuitive that I genuinely checked if I was reading the chart upside down.

For the past two years, we've all been agonising over the same question: where should I post? Instagram or TikTok? LinkedIn thought leadership or YouTube shorts? Newsletter or carrier pigeon with a Canva graphic strapped to its leg? We've been making these decisions based on human attention metrics — views, engagement rates, that dopamine hit when your reel gets 10,000 views and you briefly feel like you matter.

But here's what nobody's been asking: where does the AI look when someone asks it to recommend a product? Because increasingly, that's the question that actually determines whether your brand exists or not.

The headline number genuinely made me do a double-take. Reddit captures somewhere between 3 and 4 percent of all ChatGPT social citations. Which sounds underwhelming until you realise that's more than ten times any other social platform in the dataset.

Ten. Times.

LinkedIn's a distant second. And X — formerly Twitter, currently whatever Elon's building between rocket launches — barely registers. The platform most of us still reflexively check for "what's happening right now" is essentially invisible to ChatGPT. Whatever credibility Twitter built as the internet's breaking-news infrastructure over the past decade has apparently not transferred to AI trust. Whether that's because of the data-licensing disputes, the general post-2022 content quality situation, or because large language models just... don't vibe with the timeline anymore, the data doesn't specify. But the gap is big enough that if you're treating X as a meaningful channel for AI-mediated discovery, you might want to sit down for this.

Now here's where it gets properly interesting. Ninety-nine percent of Reddit citations point to individual discussion threads. Not subreddit homepages. Not user profiles. Not Reddit's corporate "about us" page that literally nobody has ever visited voluntarily. Individual threads.

So if you've been "doing Reddit" by seeding subreddits with product mentions and monitoring brand sentiment — that's fine for community vibes, but it's essentially invisible to the AI. What ChatGPT actually cares about is the thread itself: self-contained, genuinely answers a specific question, has real human back-and-forth. It's treating Reddit like a massive, chaotic, occasionally unhinged FAQ database. Which, honestly, is exactly what Reddit is. The AI just figured that out faster than we did.

This one broke my brain a little. When ChatGPT cites YouTube, it cites the channel page 65 percent of the time. Individual videos? 2.3 percent. Two. Point. Three.

I'll let that sink in while I stress-eat these cashews.

Channel-level citations — the main page, the /videos tab, /shorts, /community — account for nearly 80 percent of YouTube's total presence in ChatGPT responses. Which means the AI isn't watching your videos and going "wow, great B-roll." It's looking at your channel and asking "is this entity a credible authority on this topic?"

This is a complete inversion of how we've been thinking about YouTube. We obsess over individual video metrics: view counts, watch time, click-through rates on thumbnails we spent four hours designing. But ChatGPT apparently doesn't care about your individual masterpiece. It cares whether your channel as a whole reads as the definitive source on a specific subject.

So if you've published 200 product demos but your channel looks like a digital junk drawer with no consistent branding, no organised playlists, and a bio that still says "uploads coming soon!" from 2021 — congratulations, you've been optimising for humans while being completely invisible to the machines. (My partner would say this is a metaphor for my dating life before we met. They're not wrong.)

Quick caveat: this is ChatGPT-specific data. Google's Gemini has a native YouTube integration, so it likely weights individual videos differently. Which means if you're building for AI visibility across multiple models, you need to optimise for both — the channel-level authority ChatGPT wants and the video-level specificity Gemini probably favours. Easy, right? Just be perfect in two completely different ways simultaneously. No pressure.

Here's something that should make every corporate marketing team slightly uncomfortable: on LinkedIn, personal profiles get cited more than double the rate of company pages. Forty-seven percent of LinkedIn citations go to individual profiles. Company pages get 21 percent.

Let that land for a second. All those company page posts your social media manager has been carefully scheduling? The branded carousels? The "We're thrilled to announce" updates? They're generating roughly half the AI citation value of some executive's personal profile where they occasionally post about leadership lessons they learned from their golden retriever.

LinkedIn articles — the long-form publishing tool that LinkedIn has been quietly pretending is still called Pulse — account for 14 percent of citations. Not dominant, but substantial enough that if you're in B2B ecommerce and your buyers are increasingly asking AI tools "which vendor should I use for X," a well-written thought leadership piece is one of the cleaner ways to show up in that answer.

The takeaway is slightly awkward for brands that have invested heavily in corporate LinkedIn presence: the AI trusts people more than logos. Which honestly tracks. I trust people more than logos too. We all do. The AI just made it quantifiable.

Instagram is the oddball in this dataset, and I mean that as a compliment. The citation distribution is almost perfectly balanced: Reels at 36.5 percent, profiles at 30.4 percent, individual posts at 29.1 percent. No single format dominates.

Compare that to YouTube's 80 percent channel concentration or Reddit's 99 percent thread obsession. Instagram is basically the one platform where ChatGPT doesn't have a strong opinion about what format to cite. It'll pull from a Reel, a profile, or a static post depending on the query.

The optimistic read: Instagram is "resilient across models" — different AI systems with different retrieval preferences can each find something to cite. The more cynical read (and you know I live here): a platform where the AI doesn't have a clear preference might just be a platform where the AI hasn't developed strong authority signals yet. Instagram's balanced distribution could be strategic flexibility. Or it could be the AI equivalent of shrugging.

For consumer brands with visual products, the practical guidance is almost refreshingly simple: maintain content across all formats rather than going all-in on one. A profile that consistently publishes posts, Reels, and keeps an optimised bio is better positioned than one that bet everything on short-form video because someone at a conference said "Reels are the future" while wearing a headset microphone.

Here's the uncomfortable truth: we now have to optimise for two completely different audiences simultaneously. Humans, who discover products through emotional triggers, visual appeal, and that ineffable thing where a creator makes you feel like you need a new blender. And AI systems, which discover products through structured information, comprehensive answers, and source authority.

A Reddit thread from 18 months ago that thoroughly answers a question about your product category might carry more AI citation weight than your current Instagram campaign with incredible engagement metrics. A YouTube channel that's established itself as the authority on a specific niche might surface in AI responses regardless of any individual video's recent performance. Meanwhile, your viral TikTok with 2 million views might be completely invisible to every AI model on the planet.

The brands that'll navigate this best are probably the ones who stop treating "AI visibility" and "human engagement" as the same problem. They're not. They're two different games being played on overlapping fields. Building content that's structured, specific, and genuinely answers real questions is good practice regardless. But knowing which platforms and formats the AI actually reaches for when constructing responses gives that practice a direction that isn't just "post more and hope."

And look, I know what you're thinking: "Great, another thing I need to optimise for." I feel you. I really do. But the data here is genuinely useful precisely because it's specific. Reddit threads, not subreddits. YouTube channels, not videos. LinkedIn people, not logos. Instagram everything, apparently.

The way people find our products is splitting into two parallel tracks — one human, one machine — and the platforms that win in each track aren't the ones we would have guessed. Which is either terrifying or exciting, depending on how much coffee you've had.

I'm on cup four. I'm going with exciting.

Analysis based on research by Brandon Punturo and Sartaj Rajpal examining approximately 238,000 social platform citations from ChatGPT responses, October–December 2025, U.S.-based English-language users. Citation frequency data does not measure conversion value or downstream business outcomes.

Jo Lambadjieva is an entrepreneur and AI expert in the e-commerce industry. She is the founder and CEO of Amazing Wave, an agency specializing in AI-driven solutions for e-commerce businesses. With over 13 years of experience in digital marketing, agency work, and e-commerce, Joanna has established herself as a thought leader in integrating AI technologies for business growth.

😍It was great, all I needed

🙂Good, not great

🙄It sucked