5 Things About AI Search That May Surprise You

As a B2B enterprise tech marketing or communications leader, you’ve probably heard some version of this: “We need to show up in AI search.” The room nods. Slides advance. And then…someone asks how, and crickets fill the space where strategy used to be.

Two recent reports from Muck Rack’s What Is AI Reading? and Meltwater’s How LinkedIn Content Wins in AI Search give us some of the most concrete, data-backed answers we’ve seen yet. Together, they analyzed more than 34.5 million AI citations across ChatGPT, Claude, Gemini, Copilot, and Google AI. And a few of the findings? Let’s just say they rewired my brain a little.

Here’s what every B2B and enterprise tech communications and marketing leader needs to know right now.

1. AI Is Basically Ignoring Everything You Paid For

Let’s start with a stand out stat: 99% of AI citations come from non-paid media. When it comes to sponsored content, advertorials, and promoted posts, AI doesn’t care. Paid and advertorial content makes up only 0.3% of all AI citations. 

What gets cited? Media, academic sources, government sites, Wikipedia, third-party blogs, and user-generated content accounts for 84% of all AI citations. This is AI models  telling us, in no uncertain terms, that credibility is the currency. You can’t buy your way into it.

The takeaway: The investment in earned media relations, bylined articles, and third-party validation isn’t just a “nice-to-have” for reputation. It’s now table stakes for being found at all.

2. Not All AI Models Are Created Equal and Your Audience Might Be Using a Different One Than You Think

Here’s something that doesn’t get talked about enough: ChatGPT, Claude, and Gemini live in completely different information universes. The overlap in their top cited domains is almost nonexistent. They might all include Wikipedia, but after that, you’re looking at three separate realities.

ChatGPT cites sources in 96% of responses, averaging about five per response, with Axios, YouTube, and Forbes showing up consistently. Claude is the introvert of the group as it only cites sources in 55% of responses. But when it does, it goes deep, averaging 13 citations per response, pulling heavily from PubMed Central and academic sources. Gemini? Its #1 cited domain is Reddit.

Yes. Reddit. This means that the outlet getting your company covered, the platform where your executives are posting, and the content type you’re prioritizing can have radically different AI visibility outcomes.

The takeaway: Don’t optimize for one AI model. Map your audience’s likely tools, diversify your content and media coverage, and don’t assume that what works for ChatGPT works everywhere else. 

3. Your Question Matters More Than Your Answer

This one is underrated. The Muck Rack research found that the type of question asked fundamentally changes what AI cites. Industry trend queries cite journalism at 46%, which is  more than twice the rate of how-to or comparison queries. How-to questions, on the other hand, send AI reaching for reference content, owned media, and brand-authored material.

Translation: If a buyer is asking “What’s happening in [your category]?” They’re getting journalists. If they’re asking “How do I evaluate [your solution type]?” They’re more likely to land on your owned content, assuming it’s structured well.

This matters because the article you pitched to a trade pub isn’t just good for brand awareness anymore. It’s directly feeding the AI answers that show up at the top of the funnel. And the how-to guide sitting on your website? That’s working the middle.

The takeaway: Audit your content inventory through the lens of query type. Where are the gaps? Are you showing up when buyers ask trend questions? Evaluation questions? Both require different content strategies and AI search makes that painfully clear.

4. LinkedIn Is the Sleeper Hit of AI Visibility (And Most Brands Are Doing It Wrong)

The Meltwater report dropped a finding that deserves a standing ovation: LinkedIn is the #2 most cited source across B2B categories in AI search, sitting just behind YouTube (and even that ranking is inflated by Google’s dual-product setup).

But here’s the twist: 75% of LinkedIn citations come from individual member profiles, not company pages. Your brand page is not the star of this show. Your people are.

And it gets better (or worse, depending on how your executive comms program is going): the most-cited LinkedIn content isn’t polished brand copy. It’s structured, specific, answer-focused content. Think listicles, vendor comparison guides, decision frameworks, “how to choose” articles. Thought leadership pieces that lead with opinions and skip the data? Largely uncited.

Oh, and follower count? Basically irrelevant. More than 51% of citations come from members with fewer than 10,000 followers. Your quietest subject matter expert, the one who never goes viral but always knows their stuff, may be your most valuable AI visibility asset.

The takeaway: Start treating LinkedIn like the AI visibility channel it now is. Enable your subject matter experts to publish structured, data-backed content. The Meltwater report even gives you the formula: 1,500–2,500 words, clear headings, bullet lists, hard numbers, named entities. Not viral. Not flashy. Just genuinely useful.

5. Freshness Is a Feature and the Window Is Shrinking

Both reports converge on something that should change how you think about publishing cadence: AI strongly favors recent content. According to Muck Rack, 57% of media citations come from articles published in the last 12 months. The distribution peaks sharply in the first month post-publication, then drops quickly.

Meanwhile, Meltwater found that 48% of cited LinkedIn content was published within the last three months, and LinkedIn’s citation share grew 26% over just a four-week research window. Fresh content has a genuine, compound advantage.

This doesn’t mean publishing noise. It means showing up consistently with substantive content because AI models are constantly recrawling, recalibrating, and re-ranking. The brand that publishes one great piece a quarter is leaving visibility on the table that a consistent, credible publisher is picking up.

The takeaway: Build a publishing rhythm, not just a content calendar. Think monthly bylines, weekly LinkedIn posts, and a media relations program that keeps your experts visible in the outlets AI models are paying attention to.

So, What Does This Actually Mean for Comms and Marketing Leaders?

AI search rewards the same things good PR has always rewarded — credibility, third-party validation, relevance, and consistency. The difference is that now the stakes are higher, the feedback loop is faster, and the rules are more legible than they’ve ever been.

The brands that will win in AI search aren’t the ones with the biggest budgets. They’re the ones with the most useful, credible, structured, and recent voices in the market across earned media, LinkedIn, and beyond.

And if you’re not sure where your brand stands today? That’s a really good first question to answer.

Ready to see where your brand actually stands?

If this research lights a fire, good. That’s the point. But knowing the rules is only half the battle. Knowing where you stand is where strategy starts.

Given Reddit presence is increasingly a factor in AI visibility across all AI models, we are seeing that building a credible, community-native Reddit presence that AI models actually trust is one smart place to begin. 

Offleash offers a free AI Visibility Score and Audit that shows you exactly how your brand appears (or doesn’t) in AI-generated answers across the tools your buyers are using. It’s fast, human-verified, and free.

Get your free AI Visibility Score and Audit or reach out to me directly at kelly@offleashpr.com to learn more about our Reddit and AEO services.