Your customer types a question into Google. Before they see the blue links you've spent two years optimizing, they see a paragraph of AI-generated text with three or four cited sources next to it. If you're cited, you get the click and the trust. If you're not, the searcher often gets their answer and never scrolls.
That's the new shape of Google search. The AI Overview - "AIO" for short - sits above the organic results on most informational queries. SISTRIX's analysis of more than 100 million keywords shows the top organic position drops from a 27% click-through rate to 11% when an AIO is present (SISTRIX). Position #1 on the old leaderboard now competes with a box that quoted three other sites first.
Good news: the AIO cites small sites all the time. A recent analysis found 47% of AIO citations come from pages that don't even rank in the top 5 of the regular results (Wellows). The game has changed enough that a small business with the right schema, the right page structure, and a credible author byline can get cited next to Forbes and Healthline.
Here's what AIO actually weighs, and five specific tactics that move the needle.
What the AI Overview actually weighs
The AIO is not a search ranking. It's a generation step that runs after retrieval. Google pulls a candidate set of pages, then a language model decides which sentences to summarize and which sources to footnote. Four signals drive that decision.
The first is E-E-A-T - experience, expertise, authoritativeness, trustworthiness. Since the March 2026 core update, E-E-A-T scoring is the dominant signal in citation selection. The model rewards content with named authors, real credentials, and evidence the writer has done the thing they're describing. Anonymous SEO blog posts get filtered out fast. A blog by a licensed physical therapist about hip pain ranks better in the citation pool than a 2,000-word generic article from a content mill.
Second is entity recognition through structured data. The AIO needs to understand who you are before it cites you. That means schema markup - JSON-LD on your pages that tells Google "this Organization is X, this Person is Y, here's the LinkedIn, here's the Wikipedia entry." Pages with proper schema show 73% higher AIO selection rates than unmarked pages, and FAQ schema specifically lifts inclusion rates by about 60% (JigsawKraft).
Third is whether your page actually answers the question. Researchers are calling this "semantic completeness" - the share of subtopics on a query that your page covers. Pages scoring 8.5 or higher on the metric are 4.2x more likely to be cited (Wellows). Translation: thin pages that answer half the question lose to long pages that answer all of it. Length isn't the goal. Coverage is.
Fourth is source diversity. The AIO almost never cites a single source. 88% of AI Overviews cite three or more sources, and longer overviews can pull from 25-plus (Digital Applied). Google deliberately mixes a big-name source (Wikipedia, Reddit, an industry leader) with two or three smaller specialist sources. That third slot is where small businesses get cited, if your page is the most specific and best-structured answer for a long-tail variant of the query.
A fifth thing isn't a ranking signal but it matters: citations are unstable. SISTRIX measured a 54-59% citation drift rate across a 17-week window, meaning a source cited today has roughly even odds of getting swapped out within four months (SISTRIX). You don't earn a citation once and keep it. You earn it, then defend it with freshness.
1. Put FAQ schema on every service and product page
This is the single highest-leverage tactic for a small business. FAQ schema is JSON-LD code that wraps a question-and-answer block on your page and tells Google "these are common questions and direct answers." The AIO loves these because they're structured exactly the way it wants to quote.
The fix: pick five questions a real customer asks before they buy. Put those questions as H2s on the relevant service or product page, with a 2-3 sentence direct answer underneath each. Then wrap the whole block in FAQPage schema.
A skeleton you can adapt:
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [{
"@type": "Question",
"name": "How long does a physical therapy session take?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Most sessions run 45 to 60 minutes. The first visit is typically 75 minutes because we do a full evaluation."
}
}]
}
Don't stack 15 questions to game volume. Pick the five real ones. Specific beats comprehensive when the AIO is grading.
Schema markup explained: LocalBusiness, FAQPage, Product
2. Add author bylines with sameAs links to LinkedIn
Anonymous content gets thrown out of the citation pool. Google's E-E-A-T signal needs a human attached. Put a real author byline at the top of every blog post and key service page - name, credentials in plain English ("Licensed Physical Therapist, 14 years in sports rehab"), and a small headshot.
Then back the byline with Person schema that includes a sameAs array pointing to the author's professional profiles. This is what entity-resolution looks like in practice:
{
"@context": "https://schema.org",
"@type": "Person",
"name": "Dr. Jamie Chen",
"jobTitle": "Licensed Physical Therapist",
"sameAs": [
"https://www.linkedin.com/in/jamiechen-pt",
"https://www.healthgrades.com/...",
"https://twitter.com/jamiechen_pt"
]
}
The sameAs array is how Google ties your blog post to a real, verifiable professional with a public track record. Without it, the model treats the byline as a string of characters. With it, the byline is a credentialed entity. That difference moves citations.
For service businesses and clinics, add at least one industry-directory link (Healthgrades for medical, Avvo for legal, Houzz for home services). For product or e-commerce sites, LinkedIn for the founder plus the company's Crunchbase entry is the floor.
3. Structure pages as direct-answer Q&A
The AIO quotes sentences, not paragraphs. The cleaner your sentences are as standalone answers, the more citations you get.
The pattern: the question is the H2. The first 2-3 sentences underneath are the answer, written as if someone could lift them straight into a chatbot reply. Then the rest of the section can elaborate, give examples, and add context.
A bad structure for AIO:
About Our Approach
Our team has been working in this space for years and we believe in personalized care. We've found that customers respond best when we take time to understand their specific situation before recommending a solution...
A good structure for AIO:
How long does it take to see results from physical therapy for a knee injury?
Most patients with non-surgical knee injuries see meaningful pain reduction within 4-6 weeks of consistent twice-weekly sessions. Full functional recovery typically takes 8-12 weeks for moderate cases. Severe injuries or post-surgical cases can take 4-6 months.
The second version is what gets pulled into an AI Overview. The first version is what loses to it.
Aim for one direct-answer Q&A pair per H2 across your service pages and your blog. The subhead is the question. The first paragraph is the standalone answer. The rest of the section earns the trust.
Writing content that AI engines cite
4. Build topical clusters with a hub page and spoke posts
Google's AI doesn't cite a one-off blog post in isolation. It checks whether the rest of your site demonstrates depth on the topic. If you have one article on "knee pain" and nothing else, you're a tourist. If you have a hub page that defines "knee pain" plus eight spoke posts covering specific causes, treatments, and recovery timelines, you're the specialist.
The structure:
- One hub page targeting the head term ("knee pain physical therapy"). This page links out to every spoke.
- 6-10 spoke posts targeting specific long-tail variants ("ACL recovery timeline," "patellofemoral syndrome stretches," "knee pain when running downhill"). Each spoke links back to the hub and to two or three sibling spokes.
The cluster does two things at once. It tells Google you have semantic depth on the topic - the #1 citation factor. And it gives the AIO 8-10 candidate pages to choose from when someone asks any question in that cluster, instead of just one.
Pick the topic where you actually have expertise and ship the cluster there. Don't try to cluster every service. One real cluster beats five thin ones.
5. Schema-rich entity resolution at the Organization level
The Person schema in tactic #2 ties your authors to the wider web. Organization schema does the same job for your business. This is the most-skipped, highest-leverage piece of structured data on small-business sites.
In your site-wide LocalBusiness or Organization schema, add a sameAs array that links every public entity record for your company. Five categories matter:
- Knowledge graph: Wikipedia or Wikidata if you have a page. Wikidata is achievable for smaller brands than Wikipedia is.
- Professional network: your company LinkedIn page.
- Major social: Facebook, X, Instagram, YouTube where you're active.
- Business registry: Crunchbase, BBB, your state business filing if it has a public URL.
- Industry directory: Yelp, Healthgrades, Avvo, Houzz, G2 - whichever fits the vertical.
A sample Organization block:
{
"@context": "https://schema.org",
"@type": "LocalBusiness",
"name": "Your Business",
"url": "https://yourbusiness.com",
"sameAs": [
"https://en.wikipedia.org/wiki/Your_Business",
"https://www.wikidata.org/wiki/Q12345678",
"https://www.linkedin.com/company/yourbusiness",
"https://www.facebook.com/yourbusiness",
"https://twitter.com/yourbusiness",
"https://www.crunchbase.com/organization/yourbusiness",
"https://www.yelp.com/biz/yourbusiness"
]
}
The more verified entries in that array, the easier it is for Google's AI to resolve "yourbusiness.com" to a real organization with a track record. We see this rated as sameAs_completeness in audit data, scored 0-100. Sites that move from a score of 30 to a score of 80 typically see AIO citations in their cluster within 6-8 weeks.
Schema markup explained: LocalBusiness, FAQPage, Product
What to do if you're getting overviews but no citations
If you've shipped all five tactics and AIOs are still appearing for your terms without citing you, three things to check.
Freshness. Citation drift is real. Pages that haven't been updated in 12+ months drop out of the citation set even if they were strong six months ago. Add an updated dateModified and refresh the content meaningfully twice a year.
AI crawler access. Google's AIO uses Google-Extended as a separate crawler signal. If your robots.txt blocks Google-Extended, you're invisible to the model that picks AIO citations even though regular Googlebot can still index you. Check your robots.txt explicitly.
Query intent. Some queries trigger AIOs that only cite top-1% domains (Wikipedia, Reddit, Forbes, Healthline, Investopedia, NYT, big .gov, big .edu). Those 12 domains capture 47% of all citations. For head terms in YMYL verticals, you may not realistically be able to win that citation. Re-target to a more specific long-tail variant where the AIO is pulling in specialists.
The 5-tactic checklist
Pin this above your desk:
- [ ] FAQ schema on every service and product page (5 real questions per page)
- [ ] Author byline with Person schema + sameAs to LinkedIn and an industry directory on every blog post
- [ ] Direct-answer Q&A structure (question as H2, answer in next 2-3 sentences) on key pages
- [ ] One real topical cluster: a hub page plus 6-10 spoke posts on your strongest topic
- [ ] Organization schema with sameAs to Wikipedia/Wikidata, LinkedIn, social profiles, business registry, industry directory
Ship one item per week and you've changed your AIO citation profile in five weeks. The audit data we're seeing on small-business sites that complete this list shows AIO citations starting to land within 6-8 weeks of the schema shipping.
ClearGrade's AI Visibility check at https://cleargradeai.com tracks AIO citations week-over-week and ships the FAQ schema and content that closes the gaps.
Sources: - SISTRIX: AI citation drift - Wellows: AI Overview ranking factors - Digital Applied: 1,000 AI Overview citation study - JigsawKraft: schema impact on AIO inclusion