Off-Site Is the New Moat. Your Domain Is Half the Story.

There is a number from Fanout’s May 2026 B2B AEO report that should change how you draw your marketing org chart. Eighty-four percent of B2B queries on Google now return an AI Overview. That is not a forecast. That is the current floor.

The second number from the same report is the one that reorders the work. Fifty-one percent of the citations inside those AI Overviews come from off-site platforms. Reddit and YouTube combined account for more AI citations than every other external source put together. Your vendor pages are still in the index. They are no longer the dominant voice.

Two months ago AEO was a documentation discipline. This week it became a perimeter problem.

The empirical and the prescriptive landed in the same week

Fanout’s data is the empirical half. The prescriptive half arrived through Chris Long’s LinkedIn share of a Google resource titled “Build agent-friendly websites.” Google’s guidance is direct: stable layouts, clear visual hierarchy, semantic HTML, button and anchor elements over generic div containers, accessibility-tree readability. Render in a way that an agent can understand without guessing.

Read those two artifacts side by side and the message hardens. Google is telling you what to do on your site. Fanout is telling you that doing it on your site is necessary, not sufficient. The agent walks your perimeter, not your homepage.

Why off-site dominance is structural, not accidental

The 51% number is not an artifact of bad indexing. It is a direct consequence of how AI Overviews are constructed.

When a B2B buyer asks a comparison question, the model is not looking for the most polished pitch. It is looking for evidence of consensus. A vendor page is one voice. A Reddit thread with forty-seven replies is forty-seven voices. A YouTube tutorial with three thousand comments is a public lab notebook. The model treats convergence across independent surfaces as a stronger signal than any single authoritative claim, because that is how trust actually works in distributed systems.

This is the part most marketing teams have not internalized. Your message lives on your domain. The agent is reading the conversations about your message. If those conversations do not exist, you do not exist in the answer. If they exist and contradict your positioning, the answer reflects the contradiction.

The companion essays we have written cover what comes next. The 1.2M-response baseline in LLM citation attention patterns explains why the model weights certain surfaces over others. The vertical analysis in AI citation science shows the pattern is not uniform across industries. The work in platform coupling names the lock-in risk. Fanout’s report is the first one that gives B2B leaders a single number to point at in a board deck.

The two surfaces governance now has to span

AEO governance used to mean two questions. Are your pages legible to agents? Is your information architecture honest about what you do? Both questions still matter. They have been joined by two more.

The earned surface. Where is the conversation about your category actually happening? In B2B that means subreddits, YouTube channels, vendor review sites, podcast transcripts, Stack Overflow threads, niche Discord servers, and the occasional industry newsletter. None of these are properties you own. All of them feed the citation graph.

The brand-trust surface. What does the convergence of those conversations say about you? Not what your reviews say in aggregate. What the model summarizes when a buyer asks “is this vendor any good for our use case.” The answer is not under your control. The conditions for the answer being correct are.

Both of these surfaces sit outside the org chart for almost every B2B company I work with. Content marketing owns the blog. Demand gen owns the ads. PR owns the news. Nobody owns the perimeter, which means nobody is governing it.

What perimeter management looks like in practice

Translate the Fanout data into work and you get a different shape of marketing organization. Not bigger, not louder, just pointed differently.

1. Map your citation perimeter. For your top fifteen B2B queries, run them through Google AI Overviews and capture every citation. Categorize each by surface type: owned, vendor review, community, video, news, blog, other. The first time a team does this, the result is uncomfortable in a useful way. You learn that the buyer’s mental model of your category is being assembled from sources you have never sent a single email to.

2. Adopt the agent-friendly site guidance now. Google’s guidance is not optional infrastructure. Stable layouts, semantic HTML, accessibility-tree readability, button and anchor elements over div soup. These are the conditions for the agent to read your page at all. If your site fails this baseline, the off-site sources will outweigh you even more than they already do, because the model will downrank what it cannot parse.

3. Treat earned surfaces as inventory. A vendor review site that ranks in your category is not a “third-party platform.” It is a citation surface inside your perimeter. The same is true of the three subreddits where your buyers actually argue. Track them. Contribute honestly. Correct factual errors. Do not astroturf, because the model can tell, and the brand cost compounds.

4. Govern the convergence, not the messaging. You cannot control what every buyer says about you. You can control whether the things they say converge on something true. That means alignment between what your sales team promises, what your product actually does, what your docs describe, what your support replies confirm, and what your pricing page lists. When those five surfaces tell the same story, the off-site conversations converge with you. When they do not, the AI Overview surfaces the contradictions.

5. Assign the perimeter. This is the unglamorous one. If perimeter management is not on someone’s job description, the work does not get done. The role is not “AEO manager” in the narrow sense. It is closer to “brand-trust ops”: a function that owns the citation graph the way SEO once owned the link graph.

The competitive reordering is already underway

Fanout’s report contains a second-order finding that is easy to miss on first read. The companies winning B2B AI Overview citations right now are not the ones with the biggest ad budgets. They are the ones with the most coherent off-site presence: a Reddit AMA from the founder, a YouTube channel with practitioners using the product unscripted, vendor reviews that match the marketing claims, a documentation site that agents can actually parse.

None of those things are expensive in absolute terms. All of them are expensive in coordination terms, because they require marketing, product, support, and developer relations to operate as one perimeter rather than four channels. The companies that figure this out first will not be the loudest. They will be the ones whose off-site presence and on-site presence describe the same product.

Do this now

Pick your top three B2B queries. Run each through Google AI Overviews this week. Capture every citation. Tag each as owned or off-site. If your owned share is below thirty percent, you have a perimeter problem, not a content problem. The fix is not more pages on your site. It is whether the conversations the agent is already reading converge on a story you would be proud to put your name on.

The site is no longer the only governance surface. It is half the story. The other half is being written by the people who already use you, and the agent is reading every word.

This analysis synthesizes Off-Site Is the New Moat — B2B AEO Report (Fanout, May 2026) and Google’s “Build Agent-Friendly Websites” guidance via Chris Long (May 2026).

Victorino Group helps marketing and brand-trust leaders manage the AEO perimeter that now extends beyond their domain. Let’s talk.