Reading the 85% figure: what third-party data shows

Start with what was counted

The headline figure comes from an AirOps review of more than a billion AI citations. Roughly 85% of the brand mentions in those answers traced back to third-party pages, and Writer.com reported the result in late July 2026.

The same review found that brands earned citations through outside sources 6.5 times as often as through their own domains. Both numbers describe where AI answers got their information about brands, so they are a count of sources and not a verdict on any one business.

Why the size of the sample matters

A billion citations is a very large set, which makes the overall direction hard to dismiss. A pattern that holds across that many answers is hard to write off as chance.

Size has a cost, though. A share calculated across a set that large averages many different situations together, so the exact percentage works as a summary and not as a forecast for your category.

What the finding can tell you

• Off-site presence deserves a real share of your time and budget.
• A polished homepage alone rarely decides which business gets recommended.
• Review platforms, directories, discussion threads, and earned media all feed AI answers.

Treat it as a reliable direction for where to spend effort. That is a useful thing for research to give you, even without a promise attached.

What it cannot tell an individual business

A market-wide share cannot tell you your own share. Your business could sit above or below the average, and the only way to know is to measure your own visibility.

It also cannot say which outside source matters most for you, how fast your numbers will move, or whether a given tactic will produce a citation. Those answers come from your own checks, repeated over time.

Finally, a count of where citations came from is not a recipe. It shows what AI tools quoted and leaves the order of your work up to you.

Myths the figure gets stretched to support

Large numbers invite large conclusions. These are the conclusions the data does not back, along with what the wider research supports instead.

Myth: your website is now worthless

The figure says outside pages supply most brand mentions. It never says your site plays no part, because the site supports crawlability and credibility and gives third-party mentions something solid to point back to.

AI crawlers need efficient access as well. A robots.txt file that accidentally blocks legitimate AI crawlers cuts off a key visibility signal, so check yours before drawing any conclusion about what the site is worth.

Myth: one tactic guarantees a citation

The research promises no result from a single move. SE Ranking research, cited in several 2026 GEO reports, found that domains with active profiles on Google Business Profile, Yelp, Trustpilot, or Capterra are 3x more likely to be named by ChatGPT.

Read that wording closely. "More likely" describes odds across many domains, not a switch you flip, and the same holds for the finding that brands discussed often on Reddit and Quora see their chance of an AI citation run about 4x higher.

Myth: publishing more posts will close the gap

Volume on your own blog is a weak lever when most mentions come from elsewhere. One strong article on a third-party site can be worth more for AI citations than dozens of entries on your own blog.

Quantity also cannot rescue thin content. Specific, genuinely helpful pages are the ones AI systems prefer to cite, in traditional search and AI search alike.

Myth: a citation, once earned, is permanent

The research captures a moment, and citations move. They fade when content goes stale, so core pages need regular reviews and updates.

Position on the page plays a part too. AI tools lift answers from the opening of a page much more often than from its middle or end, so lead each page with a direct answer.

Myth: results should show within weeks

The data does not point to a fast payoff. Gains in AI citation frequency usually start to appear between 60 and 120 days after steady work on third-party authority and structured content begins, provided the SEO foundation is already solid.

What AppWT's own test showed

AppWT measured its own case on October 3, 2026, by asking ChatGPT and Perplexity ten questions about the company. The size question showed the limit of any single snapshot: Perplexity mentioned AppWT's team leaders and contractor network but not its AI agents, and ChatGPT mentioned none of the three.

That gap pointed to pages the engines had not read yet, which is why one reading says little about where a business will stand a few months later. The full results are here.

How to apply the research to your own business

Use the figure to set priorities, then gather your own evidence. Ask the AI tools your buyers use how they describe you, and compare the answers again after each round of work.

Buyers now use AI tools while they research options, weigh vendors, and choose, often before visiting any website. Habits such as collecting reviews, keeping directory listings current, and talking with local media may already be on your list, and the research argues for running them as one coordinated plan.

Curious how Gemini, ChatGPT, Perplexity, and Google AI Overviews describe your business right now? Contact AppWT to talk it through, or schedule a call and AppWT will walk you through an AI visibility plan for your Michigan business.