A Colorado business owner called us last month genuinely confused. He ranks #2 for his main service keyword. He’s ranked there for two years. But when he searches that term, Google’s AI answer box quotes three competitors and never mentions him. His traffic is down. His rankings haven’t moved. Both things are true at the same time, and that’s the whole story of search in 2026.
Google’s AI Overviews — the generated answer that sits above the blue links — now handles a large share of informational searches. And getting quoted inside that answer turns out to be a separate contest from ranking. Ahrefs, analyzing roughly four million AI Overview URLs, found that only about 38% of cited pages also rank in the top 10 for the same query. Ranking helps. It does not qualify you.
That leaves most business owners with a practical question nobody was answering well: what actually gets you into the answer? A recent independent study from Canadian agency HeyTony took a useful swing at it — 100 real customer-style queries across ten industries, each checked three separate times, with a control group of pages that ranked on page one and never got cited. That control group is the part that matters, and it’s the part most AI-search studies skip. If you only study winners, you’ll conclude that winners have schema markup. Compare them to the losers and you find the losers have it too.
We’ve been testing the same ideas against our own Colorado client accounts. Here’s what holds up, translated for a business that serves the Front Range rather than a national brand.
Before you change anything, understand how unstable this surface is. In the HeyTony study, 45.3% of citations showed up exactly once across three checks and never returned. Only 31.9% held across all three.
The practical translation for you: if a competitor sends you a screenshot of an AI Overview quoting them, or a tool reports you’ve “lost AI visibility,” a single snapshot is close to a coin flip. Check any citation three times on three different days before you celebrate or panic. We build this into client reporting now, because we’d rather tell you something true once a month than something exciting once a week.
This was the single strongest predictor of a stable citation in the study, and it’s the one that hits Colorado businesses hardest. Pages whose claims matched the geographic level of the AI’s answer were cited at a rate roughly 14 percentage points higher than ranking pages that got skipped.
Here’s why that matters locally. When someone searches “how much does a kitchen remodel cost” without naming a city, Google’s AI answers at the state level: “A kitchen remodel in Colorado typically costs…” A page that scopes every number to Castle Rock cannot support that sentence. It doesn’t matter how good the page is. The label on the claim disqualifies it.
Every figure on the page is scoped to one town. The AI is writing a sentence about Colorado, and this page can’t back it up.
States the Colorado-wide range first, then breaks down what Denver metro, Colorado Springs, and mountain-town projects actually run.
This is not an argument against city pages. Local pages still do the heavy lifting for map-pack visibility and for searchers who name your town. It’s an argument about your cost, timeline, and “how it works” content: lead with the state-level answer, then localize underneath it. The AI handles localization by asking the searcher follow-up questions. Your city page isn’t where that happens anymore.
This one bruises a little, because good operators hate it. Google’s AI writes its answer first, then finds pages that support the sentences it already wrote. You aren’t being judged on quality. You’re being matched on agreement.
In the study, pages whose figures matched or encompassed the AI’s figures were about 12 percentage points more likely to be cited than the control pages — and that edge grew for the citations that held across all three checks. The reverse was brutal: a genuinely expert guide built around the argument that the commonly quoted timelines are wrong was never cited once. Being right and disagreeing makes you unquotable.
The move is to encompass first, refine second. If the consensus range for your service is $8,000–$25,000 and you price at the top of it, don’t open with your number. Open with the honest full range, then add your real pricing underneath as the expert layer: “Most Colorado projects run $8,000–$25,000. Ours typically land between $18,000 and $25,000, because we don’t subcontract framing.” You keep your positioning and you become quotable.
Whether you can win a query is mostly decided before you type a sentence. The study scored every query by how much of page one belonged to big brands, banks, news sites, and aggregators. Where page one was mostly smaller sites, small and mid-size businesses took 77% of the stable citations. Where it was walled off by major brands, the big brands took 76%.
“Why is car insurance so expensive for new drivers” belongs to the national carriers and rate aggregators. No article you publish is changing that. But the more specific version of almost any question usually sits wide open, because big brands write generic. “Why does homeowners insurance cost more in wildfire zones in Colorado” is a completely different competitive picture.
The five-minute check before any post: search your target query, count the big-brand results on page one. More than 7 of 10 — pick a more specific query. Seven or fewer — you have a real shot.
“You’re not trying to get a citation. You’re trying to be the page the answer gets built from.”
In the large majority of overviews analyzed — roughly 82% — a single page supplied the backbone of the answer, and the other citations were garnish around the edges. The garnish citations were also the unstable ones. The skeleton page held its position check after check.
So the target isn’t a citation. It’s being the page the answer is built from. Practically, that means your money page needs the headline range, the tiers, the cost or timeline drivers, and the main caveat all in one place, so the AI never needs a second source. Most business sites split this across four thin pages and a blog post. That structure loses to one complete page every time.
This is the part worth reading twice, because some of it is advice we’ve all repeated.
Answer boxes, FAQ blocks, structured layouts. Around 85% of page-one results already have them, cited or not. Structure gets you ranked. It doesn’t get you picked.
Cited pages actually averaged slightly fewer structured facts than skipped ones. Volume of data points isn’t the signal. Relevance to the specific sentence being written is.
Pages with question-style titles matching the query performed slightly worse than controls. Google cites specific facts wherever they live — the study found a Terms of Service page stably cited for a legal question because it happened to hold the needed fact.
That last one points at the most useful takeaway in the whole thing: put one specific, true, concrete fact that nobody else has into everything you publish. Your own job data. Your own timelines. Your own permit-approval averages for Douglas County. A national brand can copy your formatting, your headings, and your FAQ schema in an afternoon. They cannot copy your data. A single distinctive fact can get you cited in overviews you’d otherwise never crack.
When we rebuilt Easy Home Construction’s site, we consolidated scattered service and pricing content into complete, answer-first pages that lead with Colorado-wide ranges before drilling into their service area — the same structure this research supports. Their organic visibility grew substantially, and their paid campaigns had far better pages to send traffic to.
You don’t need a new content strategy. You need about ninety minutes on your most important page.
None of this replaces the fundamentals. You still need a fast, well-structured site that ranks — being cited is a second contest, not a substitute for the first. But if you’re ranking well and still feel invisible, the gap is almost always in how your claims are scoped and how completely a single page answers the question.
We’ll review your top-performing pages against how AI Overviews actually select sources — claim scope, answer completeness, and who owns page one for your target queries — and show you exactly what to change first.
Book a Free Strategy Call →✓ No contracts ✓ Response within 24 hours ✓ 100% free
Citation stability, geographic scope, consensus agreement, page-one composition, and single-source dominance figures are from HeyTony’s June 2026 study of 100 AI Overview queries across ten industries, conducted in Canada. Top-10 overlap data from Ahrefs’ analysis of approximately four million AI Overview URLs. Snippet position data from CXL’s analysis of 100 AI Overview citations. Mountain Peak Marketing’s observations are drawn from our own Colorado client accounts.