A new study published by Search Engine Land (September 2026) tracked 775 AI citation events and found that AI search engines pull most of their answers from lists other people publish, not from the business's own website. In one 30 day test, third party listicles made up 85.8% of source mentions. The owned site got 14%. Here is how to get cited by AI search using what the data shows, written for contractors.
⚡ QUICK SUMMARY
AI engines like ChatGPT, Gemini, Claude, and Google AI Overviews are picking contractors for homeowners right now. Two structured experiments show they trust lists made by OTHER sites far more than your own pages, and they favor plain visible answers, question headings, and fresh dates. This post gives you the 5 moves that match the data, in order.
Most contractors think the way to get cited by AI is to write more pages about themselves. The new data says the opposite. AI models decide who to name based on who else is already naming you.
Hey, it's Keith. My team has worked with over 5,000 home service businesses since 2013. When a real test comes out with real numbers, I read the whole thing so you do not have to. This one changes how I would spend your next 90 days.
What did the new GEO study find?
It found that AI engines cite third party sources far more than owned content. Across two experiments, 15 commercial keywords, and 775 logged citation events, lists published on other sites drove most of the AI citations (the study, published September 14, 2026).
The author ran two tests back to back and tracked every citation by hand. Here is what each one showed.
Test 1: an established brand over several months
The team placed lists on sites the AI platforms already quoted, backed them with PR and guest posts, and tracked ChatGPT, Claude, Gemini, and Perplexity. The brand ended up showing for 10 to 12 of the 15 keywords. Lists made up 72.4% of AI citations, PR 24.1%. One single list on an outside site earned 190 mentions, more than every other source combined.
Test 2: a brand new company, 30 days, zero AI presence
This one is the eye opener. Starting from nothing, the brand got 298 AI appearances in 30 days. Of 437 source mentions, third party lists produced 85.8%, the brand's own list page produced 14.0%, and press releases produced 0.2%. First citations showed up between 1 and 18 days after a placement went live.
Why does this matter for a contractor?
Because AI search is already part of how homeowners find you. 45% of consumers now use AI tools for local business discovery, while Google still leads at 71% (BrightLocal Local Consumer Review Survey, 2026). Google AI Overviews now appear on 48% of searches, up 58% year over year (BrightEdge, 2026).
So when someone in your town asks Gemini or ChatGPT "who is a good roofer near me," the AI system is building an answer from sources it trusts. The study says those sources are mostly lists and review sites, not your homepage. If you are not on those lists, you are not in the answer. Meanwhile the contractor down the road who is on them gets named, and the homeowner never sees your name at all.
This is the same thing we saw when ChatGPT got access to 330 million Yelp reviews. AI platforms rent trust from sites that already have it. I wrote about the bigger shift in AI recommendations vs Google Maps, and this study puts numbers behind it.
How to get cited by AI search: the 5 moves the data supports
Here is the order I would work in. The first two moves come straight from the citation data. The last three come from the content findings in the same study.
1 Get on the lists AI engines already quote
Ask an AI engine your own money question: "best plumber in [your city]" or "who should I hire to replace a water heater in [town]." Write down every site it names. Those are your targets. Local news roundups, "best of" lists, trade association directories, supplier partner pages, Chamber of Commerce lists, and review sites. Getting your business added to one list the AI already reads did more in the study than every owned page combined.
2 Make sure your listings agree with each other
AI models cross check. If your name, address, phone, and services differ across Google Business Profile, Yelp, Angi, the BBB, Nextdoor, and your site, you look like two or three businesses instead of one trusted one. Fix the basics before you chase new placements. Our step by step guide to showing up in AI search walks through the cleanup.
3 Put the answer in the first 100 words, in plain view
The study found FAQ content earned more citations when it was visible on the page, not hidden in a click-to-open accordion. Answers near the top did better too. If a homeowner (or an AI crawler) has to click to see your answer, treat it as if it is not there.
4 Write headings as questions, and answer each one by itself
"How is AI search different from regular search?" beat "AI search vs regular search" in the tests. Sections that stood on their own also beat combined sections. For your site: one service question per heading, one clear answer under it. "How much does a new furnace cost in Ohio?" then the answer. No mystery, no scrolling.
5 Show a date and keep it fresh
Visible publish dates and current data correlated with stronger citation results. In the second test, half of the sources that got cited stopped being cited within 30 days. AI answers churn. A page from 2021 with no date loses to a dated page from this month, even if yours is better.
Does your own website still matter for AI citations?
Yes, but its job changed. Your site is the proof the AI checks after another source names you. Owned content produced 14% of citations in the 30 day test, and in the longer test the brand's own domain did earn citations once it had authority behind it. Think of your website as the reference page and the outside lists as the referrals.
That is why I keep saying good SEO is good GEO. Google's own Danny Sullivan said the same thing (Search Engine Journal, 2026). Clean structured data and schema markup, clear service pages, real reviews, and a fast site help traditional search engines and answer engines at the same time. You do not need a second strategy. You need the one you have to be done right, plus the outside mentions this study proves matter. If you want the basics of answer engine optimization, start there.
What does this look like for a real contractor?
Picture an HVAC contractor with 80 Google reviews and a decent site. A homeowner asks ChatGPT for the best AC repair company in their suburb. The AI names three contractors pulled from a local magazine's "top HVAC pros" list and a trade directory. Our HVAC contractor is on neither. It has better reviews than two of the three. It still loses the call. This is a composite of what we see, not one client, and results always vary.
Now run the 5 moves. The owner finds the two lists the AI quoted, gets added to both (one took an email, one took a paid membership). She rewrites her top 6 service pages with question headings and visible answers, adds dates, and matches her listings. The study says the first citations could show in 1 to 18 days. Then she asks the AI the same question every Monday and logs it. That is the whole system. Small, boring, and measurable.
Common questions about getting cited by AI
How do AI search engines decide who to cite?
AI platforms lean on sources they already trust: high authority lists, review sites, news, and directories. The study saw government, education, news, and industry specific sites outperform general sites. Domain authority of the source matters more than the polish of your homepage.
Do I need special AI crawler settings or schema markup?
Let AI crawlers in (do not block them in robots.txt) and use standard structured data for your business, services, and FAQs. That is table stakes, not a magic switch. The study did not find any technical trick that beat being on the right list.
Can I track AI search traffic?
Partly. Google Search Console does not split out Google AI Overview citations yet. You can see referral traffic from chatgpt.com, perplexity.ai, and gemini in your analytics. The simplest AI visibility audit is manual: ask 10 real customer questions in 3 AI engines once a week and record who gets named.
Is generative engine optimization different from traditional SEO?
Same foundation, different scoreboard. Traditional SEO tracks your rank. Generative engine optimization tracks how often AI generated answers name you. The work overlaps by about 80%. The new part is earning mentions on outside sources, which this study shows is the biggest lever.
What the brands AI engines cite most have in common
Look at the sources that keep winning AI citations across ChatGPT, Perplexity, Gemini, and Google's AI Mode, and the same signals show up. Use this as a checklist for your own pages and for the outside pages that mention you.
Third party mentions on trusted domains
Brand mentions on industry publications, news outlets, Reddit threads, Wikipedia style reference pages, and review sites carry more weight than anything on your own domain. Third party authority is the citation lever most brands ignore.
Content structure built for extraction
Direct answers in the first two sentences, question headings, short paragraphs, bullet points, and FAQ sections with FAQ schema. AI models pull passages, not whole pages, so structure content so one passage answers one question.
Original data and specific claims
Pages with original research, real numbers, and dated statistics get cited more than pages with vague claims. A contractor can do this: publish your average job cost, your response time, or what you learned from your last 100 estimates.
Freshness and entity consistency
Updated content with a visible date, plus the same business name, address, and category everywhere, tells the AI platforms you are one real entity that is still active. Stale content and mismatched listings lower your citation rate.
Tracked prompts, measured weekly
The brands that improve measure their AI visibility. Pick 10 customer queries, run them in 3 AI engines every week, log which competitors get named, and watch your citation share move as you add placements and fix pages.
The cost of waiting
Every month you are not on the lists, the AI keeps handing your jobs to the contractors who are. Those citations compound: the study's biggest placement stayed the top citation source for months. The brand that gets there first gets quoted first, and AI answers tend to repeat what they already said. Start now, while most contractors in your area still think this is a website problem.
| Want to Know Who AI Names Instead of You? We will ask the AI engines your customers' real questions, show you which lists they pull from, and map the placements and page fixes to get you cited.
|
About the author: Keith Carpenter is the founder of SpeedMobi, a done-for-you digital marketing company serving home service contractors since 2013. His team has built marketing systems for more than 5,000 businesses across 87+ trades.
Written for contractors, not marketers. What is working in local search and AI right now, in five minutes. No fluff, no spam, unsubscribe anytime.
Connects to QuickBooks Online in one click. Your competitors ask 100% of customers for reviews automatically. Now you can too.
See How It WorksTake the free assessment and get a full report on where your marketing stands and what to fix first.
Start My Free Assessment














