I spent three months testing what makes AI engines cite one source over another. I published identical information in different formats, tracked which versions got picked up, and measured what changed when I added schema markup or restructured content. The pattern became clear: AI search rewards businesses that make information easy to extract and verify. This isn’t about gaming algorithms. It’s about presenting your expertise in formats that AI engines can confidently cite. For Idaho businesses competing against national brands, this levels the field. A Meridian HVAC company with properly structured content beats a national chain’s generic blog post every time.
How do AI engines decide which sources to cite?
AI engines work differently than Google’s traditional search. They’re not just ranking pages by authority and links. They’re evaluating whether they can extract a reliable answer and attribute it correctly.
I’ve watched three factors determine citation:
Extractability. Can the AI pull a clean answer without ambiguity? Content that starts with “The answer is…” or “X means…” extracts cleanly. Content that circles around a topic for three paragraphs before making a point gets skipped.
Verification. Does the same information appear in multiple places? AI engines cross-reference. If your business name, address, and phone number match across your website, Google Business Profile, and industry directories, you’re verifiable. If they don’t match, you’re a question mark.
Structure. Can the AI understand what type of content this is? A blog post with proper Article schema that declares itself as an article gets treated as an article. Content without schema gets evaluated as generic text, which means lower confidence in citation.
I tested this with a Boise roofing client. We rewrote their “About Our Services” page to start with direct answers, added LocalBusiness schema, and fixed NAP inconsistencies across 12 directories. Within six weeks, they showed up in three different AI Overview results for roofing questions with “Boise” in the query.
What schema markup do I need for AI search?
Schema tells AI engines what your content represents. Without it, they’re guessing. With it, they know.
Three schema types matter most for local businesses:
LocalBusiness schema. This goes on your homepage and contact page. It declares your business name, address, phone, hours, service area, and business type. Every Idaho business should have this. When AI engines evaluate local queries (“electrician near me”, “Boise tax prep”), they look for LocalBusiness signals first.
Article schema. This goes on every blog post and resource page. It tells AI this is published content with an author, publish date, and specific topic. Articles with proper schema get cited at 3x the rate of articles without it, based on my testing across 40 client posts.
FAQPage schema. This is the secret weapon. Pages with FAQ schema get pulled into AI answers at higher rates than any other content type. The question-answer format is exactly what AI engines want. If you have a page answering common customer questions, mark it up.
I don’t write my own schema by hand anymore. I use plugins (Yoast or RankMath for WordPress) or I hire it done. The point is implementation, not perfection. A LocalBusiness schema with eight fields beats no schema. You can always add more detail later.
One warning: don’t mark up content as something it’s not. If it’s not actually a FAQ page, don’t use FAQPage schema. AI engines penalize misrepresentation harder than they reward proper markup.
What content format gets cited most often?
I’ve published the same information in five different formats to see what AI engines prefer. The winner is always the same: answer-first, hierarchically structured content with visual breaks.
Here’s the format that works:
Start with a direct answer. First paragraph, first 80 words, answer the question. No preamble. No context-setting. If someone asks “how much does it cost to replace a roof in Idaho?”, start with “Roof replacement in Idaho costs between $8,000 and $24,000 for a typical single-family home, depending on size, material, and slope.”
Use natural-language headings. Your H2s should be questions people actually ask: “What factors affect roof replacement cost?” not “Cost Factors.” AI engines match headings to queries. Question headings match more queries.
Break content into scannable chunks. Bulleted lists, short paragraphs (3-4 lines max), and occasional tables all help. AI engines parse structured content more confidently than walls of text.
Be specific with numbers and locations. “$1,200 to $1,800 per month” beats “affordable.” “Boise, Meridian, and Eagle” beats “the Treasure Valley.” Specific information gets cited because it’s useful.
Add a TL;DR section. I put this at the top of long posts. It’s a 50-80 word answer block that AI can extract without reading the full article. About 60% of AI citations I’ve tracked come from TL;DR sections or first paragraphs.
I tested this format against traditional blog structure (intro, background, then answer buried in paragraph seven) across 20 posts. Answer-first posts got cited in AI results 4-5 times more often.
What are entity signals and why do they matter?
An entity is how AI engines understand that “Boise Marketing Guy” the business is the same entity across all mentions on the web. Strong entity signals make you citable. Weak signals make you invisible.
The foundation is NAP consistency: name, address, phone number must match exactly everywhere they appear. Not “999 W Main St” on your website and “999 West Main Street” in your Google Business Profile. Exact matches.
I audit this for every client. We check:
- Website footer and contact page
- Google Business Profile
- Bing Places
- Apple Maps
- Industry directories (BBB, Chamber of Commerce, trade associations)
- Social media profiles
- Press mentions and local news citations
Inconsistencies dilute your entity. AI engines see conflicting information and lower their confidence in citing you.
Beyond NAP, citations matter. Getting mentioned in local news (Idaho Press-Tribune, Boise Dev, Idaho Business Review) creates entity signals. Being listed in industry directories creates signals. Having a Wikipedia page creates powerful signals, but that’s only realistic for larger or historically notable businesses.
For most Idaho small businesses, the practical play is:
- Fix NAP across your top 15-20 online properties
- Claim and complete your Google Business Profile fully
- Get listed in 3-5 relevant industry directories
- Publish consistent content under your business name
I’ve seen this move businesses from zero AI citations to 3-4 citations in competitive queries within 90 days. The effect compounds. Each citation makes the next one more likely.
What should I measure to know if this is working?
Traditional analytics don’t capture AI search visibility. You need different measurements.
I track four things:
Manual AI query testing. Once a week, I run 10-15 queries in ChatGPT, Claude, and Google that my business should answer. I note which queries cite me and which don’t. This is manual, but it’s the only way to see actual results. I keep a simple spreadsheet: query, date, cited yes/no, which AI engine.
Google Search Console AI Overview impressions. GSC now separates AI Overview impressions from regular search impressions. Check the Search Results Performance report, filter by search appearance, select AI Overview. This shows which queries trigger AI Overviews with your content.
Referral traffic from AI engines. Some AI citations link back. Check referral sources in Google Analytics for traffic from chatgpt.com, claude.ai, and ai.google.com. It’s small volume now, but it’s growing.
Entity coverage. Every quarter, I Google my business name in quotes (“Boise Marketing Guy”) and count distinct domains that mention me with correct NAP. Growth here predicts citation growth. I went from 23 domains to 67 domains over 18 months, and AI citations increased proportionally.
What I don’t measure: ranking position. AI search doesn’t have positions. You’re either cited or you’re not. There’s no page two.
I also don’t chase AI optimization scores from tools. Most SEO tools haven’t figured out AI search yet. Their scores reflect traditional SEO factors that don’t predict citations well.
The measurement cycle I recommend: test queries weekly, check GSC monthly, audit entity coverage quarterly. That cadence catches changes early enough to adjust course.
When will AI search actually matter for traffic?
It already matters if you’re in certain verticals. I’m seeing real traffic impact for clients in professional services (legal, accounting, consulting), home services (HVAC, plumbing, roofing), and healthcare.
Google AI Overviews now appear in 15-20% of searches, based on what I see in Search Console across 30+ client accounts. That percentage is higher for how-to queries and local service queries, lower for navigational and shopping queries.
ChatGPT and Claude don’t send much traffic yet because they don’t link consistently. But they’re being used for research before purchase. People ask AI for recommendations, then search for the businesses AI mentioned. I can’t measure that attribution, but I see it in client interviews.
The curve looks like mobile in 2011. Early adopters who optimized for mobile before it was critical built advantages that lasted years. Businesses who waited until mobile was 40% of traffic were already behind.
For Idaho businesses, the opportunity is timing. National brands haven’t figured this out yet. Their content is still optimized for 2019 Google. A Nampa home builder with answer-first content and proper schema can outrank a national franchise in AI results right now.
I’m treating this like I treated mobile: implement the basics now (schema, answer-first content, entity cleanup), measure what’s measurable, and increase investment as traffic grows. The businesses doing this now will dominate AI search in their categories within 24 months.