Molly Ploe

Need marketing help? Contact Brafton here.

If you’ve been heads-down building topic clusters and optimizing landing pages for AI Overviews and LLMs, good. Keep doing that. But I need to tell you something that might feel like a plot twist: your own website is only half the story.

The other half is happening on sites you don’t own. Reddit threads. LinkedIn posts. G2 listings. YouTube playlists. Random listicle blogs ranking “Best Content Marketing Agencies” that you’ve probably never heard of. AI models are reading all of it, weighing it, and deciding what to say about your brand.

Let’s talk about your third party citation strategy … you have one, right?

Why Off-Page Optimization Matters

Your website is not an island; your brand presence extends far beyond your domain, through directories and profiles, comments and reviews, listicles and playlists. While on-page SEO is you telling AI who you are, off-page is everyone else vouching for you. AI models value the vouching a lot more than the self-introduction because that’s how people view your brand, too.

That brings us to AI’s favorite sources for information.

Reddit accounts for a huge share of AI citations across the board. LinkedIn isn’t far behind — it’s the No. 3 most-cited domain across AI tools overall, and No. 5 for ChatGPT specifically, trailing only Wikipedia, Reddit, Forbes and Techradar. YouTube shows up constantly too: it’s the No. 2 most-cited domain behind Reddit, and the No. 1 cited source for Google’s AI Mode and AI Overviews.

So if a citation strategy starts and stops at your own domain, there’s a whole layer of AI visibility getting built without you.

Which Directories are Actually Worth Pursuing

Niche, industry-specific directories carrying real weight with AI models isn’t a surprising finding on its own. What’s more useful is knowing how to find the right ones without wasting time on the wrong approach.

At Brafton, we’ve been working to understand the role directories play in AI visibility. Using Ahrefs’ Brand Radar, we pulled the top 1,000 sources AI models cite for our brand and a few of our competitors. It turns out, sorting through that giant list wasn’t actually the most useful step; we only found 34 actionable opportunities from the mix. The rest were generally:

  • Editorial & news sites.
  • SaaS platforms and ecommerce brands.
  • Educational, reference or government websites.
  • Fan wikis and forums.
  • Social, streaming or ad-network platforms.

We found a better approach was actually typing in the same commercial questions a prospect might ask an AI tool — something like “best company for content writing” — and seeing what showed up. Those searches surfaced more relevant, actionable directory opportunities than the broad citation export did.

The lesson: don’t chase every directory that exists, and don’t stop at the big generalist names either. Platforms like Clutch, G2 and Sortlist tend to show up across most industries, but the real opportunity is almost always in the niche, industry-specific directories unique to your category — the ones a generic “top SEO tools” roundup would never think to mention. Finding those means asking AI the specific questions your prospects are asking, then working backward from what actually shows up.

Reddit as a Strategy: Can You Influence the Conversation?

Reddit is the most-cited domain by LLMs, full stop. It’s also the sixth-largest website in the world, pulling in over 53 billion monthly visits as of this year, and it ranks for almost 500 million keywords. More than 45% of U.S. internet users visit Reddit specifically to research products before buying something.

What’s interesting is that Reddit’s semantic similarity score with AI-generated answers is only around 0.53. Semantic score refers to how much of the original content AI preserves in its generated answer. In practice, that means AI models are using Reddit content constantly, but paraphrasing the words. Reddit is less about controlling the exact message and more about making sure the conversation exists at all (and that your brand shows up in it credibly).

If Reddit sounds like a good play for your brand, you need to know some ground rules:

  1. Add value before you mention your brand. Answer questions, solve problems, share something genuinely useful. The brand reference comes second, if at all.
  2. Learn the subreddit’s culture before posting in it. Some communities ban self-promotion outright; others allow it only in specific threads or only after you’ve built a posting history. There’s no universal rulebook, so you have to check each one.
  3. Build credibility before you promote anything. Consistent, authentic participation earns trust over time. Showing up only to mention a product just looks bad.

This is also just a different kind of effort than LinkedIn. It’s less “publish content” and more “show up consistently in the right conversations.” In other words, it’s a good fit for an ongoing, lightweight cadence rather than a content calendar. Still, it’ll take some work to consistently find the right opportunities to contribute.

LinkedIn as a Strategy: Can You Control the Message?

If LinkedIn is currently just the place your company reshares blog posts, I’d rethink that.

LinkedIn is the No. 2 or No. 3 most-cited domain across AI tools depending on the platform, and unlike Reddit, its semantic similarity score sits around 0.60, meaning AI answers tend to preserve most of the original meaning or phrasing rather than restructuring it. That makes LinkedIn a stronger lever for controlling the actual message, not just showing up in the conversation.

When planning out your LinkedIn posting strategy, keep a few points in mind:

  1. Educational beats promotional. Roughly 54-64% of cited LinkedIn posts are educational. Sales pitches don’t get cited nearly as often.
  2. Original thinking wins; virality not required. About 95% of cited posts are original content, not curated or reshared material. Cited posts often had only 15-25 reactions and very few comments (sometimes none). You don’t need something to go viral. You need something genuinely useful.
  3. Match the format to the depth of the question. A quick, focused question deserves a tight post in the 50- to 300-word range. A bigger, more complex topic can support a full 500- to 2,000-word article. Both lengths get cited regularly.

There’s also a role split worth planning around: Perplexity leans toward citing Company Pages, while ChatGPT Search and Google AI Mode lean toward individual creators. If you’re building out a LinkedIn strategy, you probably want both working together: the Company Page as your authoritative anchor, your executives and subject matter experts publishing original commentary alongside it.

Planning Out Your Third-Party Citation Strategy

This might all sound very complex, but in reality, getting started on your third-party citation strategy just comes down to four things:

  1. Find out where you’re already being talked about. Look at what AI cites most often for your category. Don’t assume the biggest names are the only ones that matter. Niche, specialist sources often carry more weight than you’d expect.
  2. Go where the citations already are. That might mean pursuing listings on the directories that keep coming up, pitching relevant listicle blogs for inclusion, or building out a YouTube playlist strategy if you already have a channel that’s been sitting underused.
  3. Show up authentically on Reddit. Answer real questions, follow each subreddit’s rules and build credibility before bringing your brand name into it. This is a slower, ongoing effort, not a sprint.
  4. Publish original, useful content on LinkedIn. Educational, first-person writing gets cited regularly without needing a big following behind it, and it gives you more control over how the message actually comes across.

None of this replaces the work happening on your own site. Topic clusters, optimized landing pages and strong on-page fundamentals are still the foundation you need. But if the strategy stops there, there’s a meaningful layer of AI visibility being left on the table.

A huge part of AI visibility is sitting in places you don’t control, but can absolutely influence with the right approach.