How to get cited by AI without owning the page

Sit with the size of that dataset for a second. This isn't a handful of spot-checked prompts. It's 5.1 million logged AI responses, which makes the 90/10 split one of the more statistically grounded findings in the GEO discourse this year, not a hot take extrapolated from a dozen examples.

Most GEO advice assumes the fix lives on your own website: better structure, better schema, a cleaner llms.txt. All useful. None of it touches the 90% of the answer that comes from somewhere else entirely.

That 90% figure comes from Foundation Inc.'s Ross Simmonds, published July 9, 2026, based on an analysis of 5.1 million AI responses across B2B categories. The finding: when an AI engine recommends a vendor, it's citing a third-party source, a review site, a forum thread, a comparison page someone else built, roughly nine times out of ten. Your own content is competing for the remaining slice, and most teams don't even know the split exists.

Simmonds framed the piece around a broader argument about SaaS distribution: the companies winning right now aren't necessarily the ones with the best product, they're the ones that made their product easy to find through channels they don't control. That's always been true for growth generally. What's new is how directly it maps onto AI citation behavior specifically, since the same third-party sources that drive discovery for a human buyer are, almost one for one, the sources an AI engine leans on to answer the same buyer's question a few months later.

Why 90% of B2B AI answers cite someone else

This isn't an accident of the model. It's a trust decision. An engine generating a recommendation has an obvious credibility problem if it only ever cites the vendor talking about itself. Third-party sources read as independent, which is exactly what a buyer wants when they ask an AI engine "what should I use for X" instead of typing the question into a search bar and reading ten blue links themselves. The engine is doing the same instinctive thing a smart buyer does: weighting the recommendation of someone with nothing to sell over the pitch of someone who does.

That's consistent with what we found building our own citation research: comparison and alternatives content wins the most citations of any format, and a large share of the highest-performing comparison pages aren't written by either vendor being compared. They're written by a reviewer, an analyst, or a practitioner with no stake in the outcome.

Reddit beats LinkedIn for B2B reach

The specific number that should reorder a lot of GEO budgets: Foundation's data puts Reddit at 59% penetration among B2B decision-makers, ahead of LinkedIn. Most B2B marketing teams treat Reddit as a consumer channel and LinkedIn as the serious one. The reach data doesn't support that split anymore, and neither does citation behavior: AI engines routinely surface Reddit threads as sources precisely because the discussion reads as unfiltered practitioner opinion, which is the credibility signal the model is optimizing for in the first place.

This isn't an argument to abandon LinkedIn. It's an argument that the budget split most B2B teams are running, heavy on LinkedIn content and paid, light to nonexistent on Reddit, doesn't match where either buyers or engines are actually looking. LinkedIn content is written knowing the poster's name is attached and their employer is watching; Reddit threads, particularly in niche practitioner subreddits, read as closer to an unfiltered conversation between people with nothing to sell each other. An engine trained to weight independence as a trust signal will lean toward the second every time, and the reach numbers say your buyers already are too.

That last bar is the one worth sitting with. If your entire GEO budget goes to your own domain, you're fighting over the smallest bar on the chart. The other three are where the citations actually concentrate, and they require a completely different kind of work: showing up in the conversation, not just publishing more pages.

It's worth being honest about why this is uncomfortable for most marketing organizations, beyond the budget question. Owned content is measurable and controllable in a way third-party presence never fully is. You can commission a blog post, approve it, and publish it on your own timeline. You cannot commission a Reddit thread to go your way, or guarantee a journalist covers your story the week you need it. That loss of control is exactly why most budgets default to the channel that's easiest to plan around, even when the data says the easiest channel to plan around isn't where the citations are actually concentrating.

The six sources engines already trust

We mapped this out in more detail in the only six link types AI engines actually trust: community threads, reference sites, review platforms, authoritative news, industry associations, and primary research. What Foundation's newer data adds is scale: it's not a hunch that these sources dominate, it's a 5.1-million-response dataset confirming it. The practical shift is treating your presence on these six categories as a KPI with the same seriousness as domain content output, not as a nice-to-have PR side project.

Most link-building teams still report a single metric, referring domains, and call it done. That number was built for classic search, where volume and authority correlated well enough to be a decent proxy. It doesn't tell you anything about whether you actually show up on the six source types an AI engine leans on when it's deciding who to recommend. A brand can have hundreds of referring domains and zero presence on the one Reddit thread its buyers are actually reading before they make a purchase decision. Reporting has to catch up to that gap before the budget will.

Building the third-party presence plan

Do this next

It's also worth reframing what "resourcing" means here, because it isn't primarily a dollars question. Showing up credibly on Reddit, in a comparison someone else is writing, or in a journalist's inbox takes a specific kind of practitioner time, someone who can answer a technical question honestly in a thread, or who has a genuinely interesting data point worth pitching, more than it takes ad spend. Most B2B marketing orgs are staffed almost entirely for the owned-content half of this equation and have no one whose job is explicitly the other 90%. That staffing gap is the resourcing gap the data is actually pointing to.

One caution before the close: earning a mention on a third-party source is not the same project as manufacturing one. Coordinated review campaigns, paid Reddit seeding, and journalist pitches with nothing behind them tend to get caught, by moderators, by platform trust and safety teams, and increasingly by the models themselves, which are getting better at pattern-matching manufactured consensus. The plan below works because it's slower and harder to fake than publishing another owned page, and that's exactly why it earns the trust weighting an engine is looking for.

Pull your last GEO budget line by line and count what percentage went to your own domain versus everything else. If it's not close to matching the 90/10 split the data shows, that's not a philosophical gap, it's a resourcing gap. Start with Reddit specifically: it's the highest-reach, lowest-cost third-party channel on this list, and most competitors still haven't shown up there in a way that reads as credible. Pair it with earned media work aimed at the same six trusted source types, and track third-party citation rate as its own metric starting this month, the way we built it into the third-party citation review work we ran for clients this quarter. We rebuilt exactly this kind of third-party presence plan for a B2B staffing platform that had spent two years publishing owned content into a category where the AI answer was never going to cite them for it. The domain work didn't stop. It just stopped being the whole plan.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.