Your employees = untapped AI search potential

Here’s a stat from AirOps that should make every B2B content team genuinely uncomfortable: 85% of AI citations come from third-party platforms, not brand-owned sites.

Not your blog. Not your resource hub. Not the content you've spent years building, optimizing, and promoting through every owned channel you have.

Third-party platforms. Sources you don't own and often can't control.

Now pair it with this finding from Moz’s analysis of 40K queries: 88% of Google AI Mode citations aren't even in the organic top 10 search results.

The last decade of SEO content you’ve been focusing on? Not super relevant (especially if you haven’t been updating your top-ranking pieces.)

The B2B content model of the past (optimize for keywords, publish to the company blog, distribute across brand social channels, build domain authority over time) doesn’t copy/paste over to optimizing for LLM citations.

So what does? Turns out, Google accidentally told us.

The Signal Google Didn't Mean to Reveal

In May 2024, Google's internal ranking API was leaked. It was a significant moment that was widely covered in SEO circles, but most of the conversation focused on the wrong part.

Buried inside the leak was a signal called the Author Reputation Score, a direct input into their quality evaluation models. Alongside it were author and publisher flags that affect confidence levels in how content is weighted, and "effort", which is an LLM-based estimation of how much genuine work and expertise went into producing a piece.

Google isn't just evaluating content anymore; it's evaluating people.

The model is asking, "Is the person attached to this material a credible source?"

This isn’t just unique to Google, either. AI answer engines like ChatGPT, Perplexity, and Claude do the same thing, using what researchers call entity embeddings. Using systems thinking, the LLMs are building internal representations of individuals, not just brands or domains.

So when someone asks a question, these systems work to identify which named people are credibly and verifiably associated with a given topic, and then draw on those people's signal footprints to construct an answer.

Again, this is what I have been referring to as “source signals.”

The AirOps data makes this concrete: pages with visible author credentials are 41% more likely to be cited by AI.

Author reputation is a citation input with a name and an actual variable in Google's codebase. You better believe that’s true for other LLMs, too.

How AI Engines Decide Who's Credible

Let me walk through how this actually works, because understanding the mechanics changes how you think about your content strategy.

When someone asks an AI system, "Who's an expert in B2B content strategy?" or "What's the best approach to demand generation for SaaS?" the model is trying to determine which named humans, organizations, and concepts are credibly and independently associated with those topics.

This is called entity resolution: taking a question and figuring out which real-world people and sources can reliably answer it.

The key word there is "independently." AI systems cross-verify expertise across sources, it doesn't expect to be coordinated. It's essentially asking: Does this person appear as a credible voice on this topic in places that have no incentive to say nice things about them?

That cross-verification has a measurable effect. DigitalBloom research shows brands and people mentioned positively across 4+ non-affiliated platforms are 2.8x more likely to appear in ChatGPT responses. Earned media is also a dominant source of citations: Muck Rack found that 85%+ of AI-cited links come from earned coverage, not owned content or paid placements.

In practice, here’s how this plays out:

Say at your company, you have a product manager with hot takes and a solid LinkedIn presence. They get quoted in a trade publication, appear on an industry podcast, and contribute to a relevant online community, like a Reddit thread. There are four independent trust signals AI can cross-verify, that from different sources, with different editorial standards, none of which have an obvious reason to vouch for this person.

That's exactly what AI is looking for.

The SME who only ghostwrites for the company blog (even an exceptional, highly trafficked one) = zero independent signals. Their name might not even be on the content. And even if it is, the signal comes from one source: you, the brand. The most interested, least independent party possible.

This brings me back to, you guessed it, source signals.

A source signal is any independent, verifiable instance of a person's expertise that exists outside brand-owned channels. These include things like:

• A podcast appearance
• A quote in an industry publication is a source signal
• A LinkedIn post that gets engagement and commentary from other practitioners

A ghostwritten blog post on your company site is not a source signal.

Source signals are a new unit of currency in AI search, and they determine whether your experts are resolvable entities that help your AEO efforts (or stay invisible assets that simply tick a content to-do list box).

Your buyers are getting answers from ChatGPT. Are you in them?

85% of brand mentions in AI search come from third-party sources, not your own site. Pages refreshed within 90 days are 3x more likely to be cited. AirOps analyzed 15M+ AI queries and turned the findings into a 4-play framework used by teams at Carta, Ramp, and Webflow to earn AI citations, visibility, and pipeline.

​Get the 2026 AEO Playbook from our sponsor this week, AirOps →​

LinkedIn Is Now AI's Primary Source for Professional Answers

In March 2026, SEMrush analyzed 325,000 unique prompts across ChatGPT Search, Google AI Mode, and Perplexity, identifying 89,000 unique LinkedIn URLs cited in AI-generated responses.

Their finding: LinkedIn is the second-most-cited domain across all three platforms, trailing only Reddit. On average, 11% of all AI responses include a LinkedIn citation, ahead of Wikipedia, YouTube, and every major news publisher.

Separately, Profound tracked 1.4 million citations across six AI models from November 2025 through February 2026.

Their finding: LinkedIn is the #1 most-cited domain for professional queries across ChatGPT, Google AI Mode, Google AI Overviews, Microsoft Copilot, and Perplexity…and it got there fast.

LinkedIn's domain rank on ChatGPT sat at approximately #11 in November 2025, but by February 2026, it had climbed to approximately #5, more than doubling its citation frequency in three months, which is the largest single domain authority shift Profound observed all year.

​Spotlight's database shows the acceleration from a different angle: ChatGPT is now citing LinkedIn 4.2x more than it was a year ago. Perplexity is citing it 5.7x more. The average across all LLMs is 4 to 5 times the previous rate.

This platform, which most B2B brands have treated as a social media afterthought (see also: a distribution channel for their blog posts and/or a place for executives to post inspirational quotes), is now one of the primary places AI goes to find professional expertise.

Now, to validate the source signal argument: the type of LinkedIn content being cited is shifting just as fast as the volume.

Profound’s study noted that between November 2025 and February 2026, citations to LinkedIn profile pages fell sharply, from 33.9% to 14.5% of all LinkedIn citations. Meanwhile, citations to published content (posts, long-form articles, and newsletters) grew from 26.9% to 34.9%.

AI tools are increasingly citing the content people and companies create on LinkedIn, not just profiles.

This means that merely having a LinkedIn profile isn’t enough; what AI rewards is published expertise. Things like original analysis, named frameworks, and unique, data-informed perspectives that can be attributed to specific humans.

The Individual vs. Brand Distinction on LinkedIn

Profound also found that on ChatGPT and Google AI Mode, the two highest-converting platforms for professional queries, 59% of cited LinkedIn content comes from individual members, not company pages.

A few more specifics from the data that matter for how you think about content:

• LinkedIn articles between 500 and 2,000 words account for 72–77% of AI citations. Not short posts, not company updates, but substantive long-form pieces authored by named individuals.
• 95% of cited LinkedIn content is original. Reshared posts almost never appear in AI-generated responses.
• Creators with fewer than 500 followers are cited at a similar frequency to larger accounts. Credibility and specificity of expertise matter more than audience size.

There is one important nuance to hold here, though:

The LLM you’re tracking citations within matters. Perplexity, for example, goes the other direction. On that platform, 59% of LinkedIn citations come from company pages, not individuals, which means a complete strategy needs both: the individual employee SME content for efforts within ChatGPT and Google AI Mode, and then a company page content for Perplexity.

LinkedIn Thought Leadership Isn’t Just for CEOs

A few years ago, we all witnessed the wave of CEOs who leaned into LinkedIn thought leadership. But now, it’s time to expand those efforts beyond the Founder or face of the company and to incorporate other company subject matter experts into that mix.

The thing is: CMI data indicates 96% of B2B companies create thought leadership content. Nearly every B2B company has a blog, a LinkedIn page, an executive who occasionally publishes, and a content team working hard.

But only 37% have employees with specialized knowledge contributing to those efforts. And of that 37%, fewer than 5% of the company’s total employee roster are doing this type of work.

CMI's take on this is worth quoting directly:

"If fewer than 5% of your employees with specialized knowledge are involved, you don't have a thought leadership program; you have a content team trying to look smart on LinkedIn."

Why Mobilize Employee SMEs on LinkedIn for AEO Efforts?

Most B2B brands have essentially one to three entity-generating source signals: The brand LinkedIn page, the CEO who occasionally does a podcast, and the company blog that publishes without a bylined author.

That approach limits potential source signals in a game that rewards a breadth of independent verification.

Meanwhile, every SME inside your organization who isn't publicly sharing expertise is a source signal that isn’t being utilized.

They might be the most credentialed person in the industry on their topic, have unique insights on the pain points they’re solving for that no one else has, and even have a contrarian take or two that go against the status quo within your space (hello, LinkedIn comments section blowing up!)

Creator Economy expert Lia Haberman also makes a strong case for employee-driven LinkedIn efforts, saying:

But if they’re not authoring bylined content, sharing the experiments they’re running (or, if they’re really ambitious, creating videos about their expertise), they’re essentially invisible to LLMs.

That’s a problem with an easy solution. Let me give you a peek at the system I’ve developed to solve for this.

The Systematic Approach to Developing Employee Source Signals on LinkedIn

I want to make this concrete with an example of how this approach is different from the “old way” of leveraging insights from internal SMEs.

The old model

Process: Your content team identifies a topic. They schedule time with an internal SME like a product manager, a solutions engineer, a VP of Customer Success, and then interview them for 30 minutes. The content team writes the piece, the SME reviews it, and it goes up on the company blog under a brand byline or, generously, with a brief author bio at the bottom. The brand promotes it through official channels.

Result: one owned-media signal. The SME's expertise resides within a brand property, aligned with a generalized brand voice, discovered through a brand channel. Zero source signals for the SME as an individual, no cross-verification, no entity resolution. The next time an LLM gets asked a question this SME could beautifully answer, it won't know they exist.

The new model

Process: The SME publishes a genuine point of view on LinkedIn under their own name. Not polished company-speak, but their actual take, in the long-form article format that accounts for 72–77% of AI-cited LinkedIn content. A trade publication quotes them in a story about the same topic. They appear on an industry podcast, talking through the same ideas in their own words. The company blog publishes a piece that cites them as a source, just like a journalist would.

Result: 4+ independent source signals. The SME becomes a resolvable entity. They exist, credibly and verifiably, on the open web, including on the platform from which 11% of all LLM responses now draw a citation. When a tool like ChatGPT is asked a question this expert can answer, it finds them (and, by extension, your brand).

The infrastructure to make this work is already in place, and I've got the process to make it low-effort and high-impact.

Look at the employee advocacy data alongside the LinkedIn citation data, and the picture is unusually clear:

• Content shared by employees gets 561% more reach and is reshared 24x more than content shared by brand channels (source)
• 59% of employee advocates are already sharing original content, not just company posts, but their own perspectives (source)
• 75% of the LinkedIn authors whose content gets cited by AI post at least 5 times per month (source)

Most B2B companies view employee advocacy as a distribution tactic to generally amplify organic brand reach. That's fine, but the real value of this work lies in its citation infrastructure.

Every time an employee publishes original content under their own name, in their own voice, on LinkedIn, they're creating a source signal that AI can find, verify, and reference independently of your brand. That post, if it's substantive, specific, and consistent with their named area of expertise, has a 95% better chance of being cited by AI than a reshare of your company's content would. It's social media and AEO strategy.

The companies that figure this out will have something that's very hard to replicate: a distributed network of credible, independently verifiable human experts that AI can resolve by name.

Why? Because their people are genuinely, publicly present in the conversations that matter on a platform that AI now treats as a primary source for professional answers to queries.

The Objections + Getting Internal SMEs on Board

I've shared this concept with enough marketers to know what happens next. Someone in the room raises a hand and says, "Okay, but we already have a thought leadership program."

If you’re producing content that lives primarily on your owned channels under a brand byline, you have a content program. That's not the same thing.

The distinction here is that a thought leadership program that doesn't generate independent, cross-platform source signals for named individuals isn't building AI visibility; it’s merely distributing your blog.

Run this audit: search for your top three internal experts by name across ChatGPT and Perplexity, and identify five queries they should be answering. If they don't appear, the program isn't working for AEO, regardless of how many posts it's producing.

Other objections I’d like to squash

"Our SMEs don't have time for this."

This is almost always a proxy for something else: our SMEs don't want to stare at a blank page and figure out what to say. Those are solvable problems with different solutions. The time problem is solved by my expertise extraction model: a structured 20-minute interview once a month, with the editorial team handling all production. The SME's job is to think out loud, not to write. If the real problem is the blank page, that's actually the core thing this approach is designed to remove.

"We can build this internally."

Maybe. But the specific skill set required (structured interview technique that extracts genuine POVs rather than press releases, editorial judgment about what's citable versus what's generic, and AEO measurement across platforms) is not standard content team territory.

Most content teams can produce posts. Very few can run a diagnostic interview that surfaces the contrarian take an SME has been sitting on for three years, name it, and turn it into a citable framework in four weeks. That gap is where the work happens.

"How do we know this will move pipeline?"

Honestly? The direct causal data is still being written. What exists is powerful adjacent evidence: 73% of B2B buyers using AI in their research process, findings that 95% of deals are won by the vendor already on the shortlist, and LinkedIn citation data showing that individual voices generate 59% of AI citations on the most commercially relevant platforms. The mechanism is sound, and the brands that build this now will own the category when it does.

Also, take this insight from Stacker Media CEO Noah Greenberg, who leans into this approach wholeheartedly:

"Our employees won't do it even if we ask."

If the program is framed as "please post more content for the company," they won't. If it's framed as "we're going to make you one of the most cited experts in your domain in 18 months, and it will only cost you 10 minutes a week," the conversation changes.

The employees' and the company's interests are genuinely aligned here. They get a professional reputation that follows them wherever they work, and the company gets citation infrastructure. The framing matters, and leadership endorsement matters even more.

Companies where the C-suite is visibly participating often see a higher employee participation in advocacy programs. This is not a program you can delegate down and expect to succeed.

Employee Expertise Should Be Part of Your AI Search Efforts

LinkedIn has just handed you one of the clearest data points in years: the platform is now the #1 cited source for professional queries across every major AI system, individual voices outperform company pages on the platforms that matter most, and the content being cited is original, specific, and attributed to named humans…not reshared brand content.

My prediction: The brands that will master AI citation rate in 2026 and beyond will be the ones who build a network of human entities that AI can actually find and trust.

Here's the question I want you to answer honestly this week: How many of your SMEs could AI identify as a credible source right now?

If you have three to five, we should talk.

I’ve developed a comprehensive, systematic approach to make this a low-effort initiative that requires minimal time and effort from your already busy SMEs.

I’m currently testing it in a pilot program with two different B2B brands, but I want to open it up to a cohort of marketers or growth-minded leaders who read Context Window and want to run this process with their own organizations to see its impact on LLM citation rates.

Those who join will get access to all of my process docs that bring this program to life, direct access to me for Q&A/troubleshooting, and will be part of a small community where we share ideas, wins, etc., as we run this initiative over a 90-day sprint.

If you want to be early to the party with me on this, apply to join the kickoff cohort here.​

I’m capping the group for this initial run to keep it intimate (and because I’m giving my focused time + attention to participants, which simply does not scale!)

Not quite ready? That's okay! I’ll be sharing some high-level ideas around this premise in the newsletter over the next few weeks, and this process + framework will eventually be available for purchase (albeit at a much higher price point) after this initial cohort runs.

One last thing.

The early results for this framework are proving incredibly promising.

I’ve been running it on my own personal LinkedIn account since February, and the charts give you a peek, at least on the LinkedIn front, of the opportunity to be had here. Now, on top of that, add in the lift within LLM visibility...more on that soon.

Found this content interesting? Feel free to share it, save it to read later, forward it, or tell me why you disagree with the thesis outlined here (hit reply!)

'Til next time,