How to Get Your Medical Practice Cited in Google AI Overviews

To be cited in a Google AI Overview, a page must be indexed and eligible to appear in Google Search with a snippet. Google states there are no additional technical requirements. Selection then depends on answering a specific patient question directly, with verifiable clinical authorship and content that exists nowhere else.

TL;DR

• Google publishes one eligibility rule: the page must be indexed and snippet-eligible. There is no separate AI standard [1]
• Audit nosnippet, data-nosnippet, max-snippet, and noindex before rewriting a single page. A directive set years ago can exclude you today [3]
• Health queries of seven words or more return an AI Overview 73.9% of the time, against 35.8% for one and two word queries [5]
• Google says no special schema.org markup is required for AI features. Keep structured data anyway, for rich results and machine understanding [1][2]
• Search Console has reported AI Overview and AI Mode impressions since June 3, 2026. No clicks, no query data [4]
• llms.txt, content chunking, AI-specific rewriting, and bought mentions do nothing for Google AI Overviews [2]

One liner: Eligibility is a technical property of the page and selection is an editorial one, which is why the fix usually starts in your robots directives and ends with whose name is on the article.

Key Numbers

What a Citation in an AI Overview Actually Is

In short: A citation is a supporting link Google attaches to an AI-generated answer. It is awarded per answer, not per keyword position, which is why one page can earn citations across questions you never targeted.

When someone searches a health question, Google may generate an AI Overview above the classic results. The answer is grounded in pages pulled live from Google’s Search index through retrieval-augmented generation, and the links shown beneath it are the pages that supported the response [1]. Those links are citations.

The mechanism that matters here is query fan-out. Google issues several related searches at once to build one answer [1]. A patient asking about knee replacement recovery sets off parallel queries on recovery timelines, physical therapy schedules, driving restrictions, and pain management. Google says fan-out lets it surface a wider and more diverse set of links than a classic search does [1]. For a practice competing against NIH and Mayo Clinic on the head term, that breadth is the opening.

Two things to set expectations. AI Overviews appear only when Google’s systems judge them additive to classic Search, so plenty of queries never trigger one [1]. And the discipline has a name problem. You will see this work sold as GEO or AEO, but Google’s position is that optimizing for generative AI search is optimizing for search, which is still SEO [2]. Our generative engine optimization for healthcare guide covers that framing in depth.

Coverage is uneven, and unevenly useful to you. Reporting from healthcare search tracking puts near-zero AI Overview presence on mental health crisis queries, eating disorder searches, and addiction topics [6]. Google says it applies extra care to health content and does not publish a suppression list. If your practice works in behavioral health or addiction medicine, everything that follows applies to a smaller share of your queries than it would for an orthopedic group. Verify against your own results before committing a budget.

This article is about Google specifically. ChatGPT, Perplexity, and Gemini select sources differently, and the FAQ addresses them.

The Only Eligibility Rule Google Publishes

In short: To be eligible as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to appear in Google Search with a snippet. Google states there are no additional technical requirements [1].

That sentence does more work than the rest of the advice on this topic combined. Read what it does not contain. No schema requirement. No word count. No author bio standard. No content format. The gate is technical, and it is low.

The practical consequence is a two-stage model worth carrying through the rest of this guide. Eligibility decides whether your page can be pulled at all. Content quality decides whether Google picks yours from the eligible pool. Most published advice collapses these into one checklist, which is why practices spend months on markup while a snippet directive quietly keeps them out of the running.

Google attaches one caveat to its own rule, and it deserves repeating. Meeting every requirement and following every policy does not mean Google will crawl, index, or serve a page. Indexing and serving are not guaranteed [1]. Anyone promising citations is selling something Google declines to promise.

There is a second condition worth checking. Google’s generative AI optimization guide adds that a site must be included in Search generative AI features in Search Console to be eligible for display in those features [2]. A control to exclude a site from AI features exists in Search Console, so confirm nobody on your team switched it off during an earlier debate about AI and your content. That single toggle overrides everything else in this article.

Does Page-One Ranking Matter?

In short: Every published dataset says ranking helps. None of them agree on how much, and Google does not publish the relationship at all.

Two pages ranking for this topic right now make opposite claims. One cites an analysis of 2,400 citations finding that more than 76% of AI Overview citations come from page-one organic results [10]. Another reports that only about 17% of cited sources also rank in the organic top ten, which would put five of six citations on pages nowhere near page one [6].

Both cannot be right. They sampled different queries at different times against different definitions of a citation, and neither Google’s AI features documentation nor its generative AI guide states a ranking-to-citation relationship [1][2].

Here is what survives the disagreement. Strong organic performance raises your baseline, because AI features draw from the same index and the same ranking systems as classic Search [2]. Here is what does not survive: any specific percentage quoted as settled. When a vendor leads with one, ask which study, which queries, and which month.

Step 1: Audit Your Snippet Controls First

In short: Four directives can remove a page from AI Overview eligibility, and three of them are ones a developer might have added for reasons that made sense at the time. Check these before rewriting anything.

Google’s control for what appears from your pages in Search is the same set of preview controls that has existed for years [1].

max-snippet is the one that catches people. A value set years ago under a different content strategy still applies today, and a page capped at 50 characters has almost nothing to offer a generated answer.

Google publishes a troubleshooting sequence when content keeps appearing after a control is added, and it works in reverse for this audit. Use the URL Inspection tool to see the HTML Googlebot actually received, since a directive that is broken or invisible to the crawler is not doing what you think [3]. Then allow time. Google says crawling can take anywhere from several days to several months depending on how often its systems decide a page needs refreshing, though you can request a recrawl [3].

One more place to look, and it is specific to this industry. Google lists ensuring crawling is allowed in robots.txt and by any CDN or hosting infrastructure [1]. Medical sites carry heavier security layers than most, and a firewall rule written to stop scrapers can stop Googlebot with it. Confirm the crawler reaches your pages at the network level, not only in your robots.txt file.

Separately, Google-Extended governs AI training and grounding in some of Google’s other systems, not eligibility for AI Overviews in Search [1]. Blocking it does not remove you from AI Overviews, and unblocking it does not add you.

Step 2: Answer the Long Question in the Patient’s Words

In short: Health queries of seven words or more return an AI Overview 73.9% of the time, against 35.8% for one and two word queries [5]. The long, specific patient question is where the exposure sits, and where a practice can compete.

That finding comes from an analysis of more than 130,000 health-related queries published in 2025 [5]. The same study puts AI Overviews on 51% of health queries overall. A separate industry analysis places healthcare closer to 88% [6]. The two numbers come from different query sets sampled at different times, and neither is settled. What both agree on is direction: healthcare draws AI Overviews at a higher rate than almost any other category, and specificity raises the odds rather than lowering them.

The obvious response is to build a page for every question variant. Google closes that door. It warns that creating separate content for every possible variation of how people search, including fan-out queries, can violate its scaled content abuse policy, and states that a high quantity of pages does not make a site higher quality or more relevant [2].

So the instruction is narrower than it first appears. Build fewer pages, each owning a question a patient would actually type, and answer it in the opening one or two sentences of the section that owns it. Not after the clinic history. Not after the credentialing paragraph.

There is a number behind that instruction. A study of 100 pages cited in AI Overviews found 55% of citations came from the top 30% of the page [11]. One study, small sample, so hold it loosely. It points the same direction as the advice: an answer buried at word 400 is the answer Google reads last.

Compare “dermatologist” against “do I need a referral to see a dermatologist in Ohio.” The first is a category term you will lose to Healthline and WebMD. The second is a question with a definite answer that your front desk gives on the phone several times a week.

That gap is the whole opportunity. Cited health domains skew institutional, with NIH leading at roughly 39% of AI Overview citations, Healthline near 15%, and Mayo Clinic near 14.8% [7]. A practice does not out-authority those sites on the category term. It wins on specificity. Our AI search optimization for healthcare service page covers how we map those questions to pages.

Step 3: Make Clinical Authorship Verifiable

In short: E-E-A-T is a quality framework, not a field you fill in. What moves a health page is content only your practice could have written, attached to a clinician a reader can verify.

Google is direct about what separates content worth citing from content worth skipping. A first-hand account offers a perspective grounded in personal experience, while a summary of existing material restates what is already available elsewhere, and Google asks creators to bring their own in-depth experience to the page [2]. It draws the same line a second way, contrasting commodity content built on common knowledge against non-commodity content that offers an expert or experienced take beyond the ordinary [2].

Applied to a practice, that distinction is sharp. “What is a root canal” is commodity. Every health site has it and yours will not win it. “What we tell patients who ask whether a root canal hurts more than an extraction” is not commodity, because it comes from a chair in your office and it exists nowhere else.

You already hold this material. It is in the questions patients ask during intake, the concerns that surface at the two-week follow-up, and the referral quirks specific to your county.

Attach it to someone real. A verifiable byline needs:

• A named clinician with credentials stated, not implied
• A bio page that exists and loads, with board certifications and affiliations
• A visible review date, and a named medical reviewer when a writer drafted the page
• Author details that match the practice’s listings elsewhere, since inconsistency reads as noise

Keep the pages current. Competitors sell freshness as a lever unique to AI search, and Google’s AI documentation does not list it as one. What Google does say is that its AI features run on the same core ranking systems as classic Search [2], where recency has mattered for years. Review dates on clinical content also serve the patient reading them, which is the better reason to maintain them.

One correction worth making, because the advice circulating on this contradicts it. E-E-A-T is the framework Google’s quality raters apply and its systems aim to reward [8]. It is not a schema property and it is not something markup transmits. You cannot declare a page trustworthy in JSON-LD. The credentials, the review process, and the bio page are the signal; structured data describes what is already visible, which is the subject of the next step.

One more correction while we are here. You will read that Google’s guidelines require medical content to be written or reviewed by a credentialed professional. They do not. The Search Quality Rater Guidelines describe how human raters evaluate search results, and raters do not rank pages. Credentialed authorship is strong practice for patients and for trust. It is not a published requirement, and stating it as one makes the advice easy to dismiss.

Health content also sits in the category Google treats with the most caution, which raises the standard rather than changing the mechanism.

Step 4: Use Schema for What It Is Good For

In short: Google states there is no special schema.org structured data needed to appear in AI features [1]. Keep your markup anyway, for reasons that have nothing to do with citations.

That correction matters because the advice pointed at medical practices says the opposite. Implement these four schema types, the guidance runs, and AI will extract your answer. Google’s documentation contradicts it in two separate places. The AI features guide says you do not need to create new machine-readable files or markup, and that there is no special schema.org structured data to add [1]. The generative AI optimization guide lists overfocusing on structured data among the things site owners can ignore, and states outright that structured data is not required for generative AI search [2].

Now the other half, because this is not an argument for deleting your JSON-LD. Google recommends continuing to use structured data as part of your overall SEO, since it helps its systems understand your content and supports eligibility for rich results [2]. Rich results still shape classic search, and classic search runs on the same index that grounds AI Overviews. The markup earns its place. It sits one step removed from the outcome you were sold.

One rule governs all of it. Google lists making sure structured data matches the visible text on the page among its best practices for AI features [1], and its general structured data guidelines say the same [9]. Marking up content a reader cannot see violates its guidelines, and it is the most common error on practice sites, where FAQ markup routinely carries questions that appear nowhere on the page.

This article practices that. It ships with BlogPosting, HowTo, and FAQPage markup whose questions match the visible FAQ word for word, built for machine understanding rather than for a citation. Validate yours in the Rich Results Test before it goes live.

Step 5: Measure in Search Console, Not in a Dashboard

In short: Google launched a generative AI performance report in Search Console on June 3, 2026, and rolled the associated controls out worldwide on August 31, 2026 [4]. Most advice on this topic still says AI visibility cannot be measured. That was true when it was written.

Open Search Console, go to Performance, and look for the Generative AI entry in the left menu. Discover has its own. Here is what the report gives you and what it holds back [4]:

• Reports: impressions, pages, countries, devices, and dates
• Withholds: clicks, click-through rate, and query-level data

That combination answers one question well and another not at all. You can establish whether your pages are surfacing inside AI Overviews and AI Mode. You cannot yet tell what that visibility earns you, or which patient questions produced it.

Even so, it settles an argument practices have been having with their own analytics. A clinic watching impressions hold steady while clicks slide has usually filed that under ranking trouble. Often it is fan-out exposure: your page supporting answers across a spread of related questions, with the answer delivered on the results page. Before June, that was a hypothesis. Now it is something you can check.

A caution on the alternatives, and it comes from Google rather than from us. Google advises wariness toward third-party tools that promise ranking success or claim to use internal Google metrics, and states that no third-party tool has access to its internal ranking or AI systems [2]. Google allows that such tools can still suit a workflow, provided their advice gets evaluated against official guidance [2]. Treat vendor dashboards as directional, and treat Search Console as the record.

The report is labeled beta and its history runs back only months. A practice that starts logging now holds a baseline that a practice starting next year cannot reconstruct. That is the argument for checking your own AI visibility data this week rather than next quarter.

Local Searches Follow Different Rules

In short: “Near me” health queries appear to sit outside AI Overviews, which leaves the map pack and your Google Business Profile as the channel for patients ready to book.

Tracking research reported across the healthcare SEO trade press found that Google tested AI Overviews on local provider searches and then removed them, with local health queries showing no AI Overview presence [6]. Google’s documentation neither confirms nor denies this, so treat it as a third-party observation rather than a published rule, and check your own priority queries before building a plan on it.

The strategic half holds either way. Informational intent and local intent behave differently, and they always have. “How long does Mohs surgery take” is a question an AI Overview can answer. “Mohs surgeon near me” is a decision that ends in a phone call. The first competes for a citation. The second competes on the map pack, on review volume, and on whether your Google Business Profile is accurate.

Practices that fold both into one plan tend to underinvest in whichever half their agency finds less interesting.

What Google Says You Can Ignore

In short: Four tactics sold as AI visibility work are ones Google’s documentation states it does not use. Each line below paraphrases Google’s own wording so you can check it against the source [2].

A fifth belongs with them, covered back in Step 2: building a separate page for every question variant runs into Google’s scaled content abuse policy, and page volume does not raise site quality [2].

One limit on all of this. These four statements describe Google Search. ChatGPT, Perplexity, and Claude select and cite sources through different systems, and some of them do read files that Google ignores. The accurate conclusion is narrow: llms.txt does nothing for your Google AI Overview citations. It is not a verdict on the file everywhere.

Which brings the argument back to where it started. Google’s own position is that optimizing for generative AI search is optimizing for search, and remains SEO [2]. There is no separate discipline to buy. The work that earns a citation is the work that earned rankings all along, done with sharper attention to which question the page answers, how quickly it answers it, and whose name is on it.

Frequently Asked Questions

Does schema markup get my practice cited in AI Overviews?

No. Google states there is no special schema.org structured data needed to appear in AI features, and lists overfocusing on structured data among the things site owners can ignore. Structured data still earns its place for rich-result eligibility and for helping Google understand your content, so keep it. It sits one step removed from the citation itself.

What are the requirements to appear in a Google AI Overview?

A page must be indexed and eligible to be shown in Google Search with a snippet. Google states there are no additional technical requirements. Meeting them does not guarantee anything, since Google also says indexing and serving are never guaranteed.

Can I see whether my pages appear in AI Overviews?

Yes, since June 3, 2026. Search Console has a generative AI performance report under Performance that shows impressions, pages, countries, devices, and dates for AI Overviews and AI Mode. It does not show clicks, click-through rate, or which queries produced the impressions.

Does blocking Google-Extended remove me from AI Overviews?

No. Google-Extended governs AI training and grounding in some of Google’s other systems, not eligibility for AI Overviews in Google Search. Blocking it does not remove you and unblocking it does not add you. The controls that affect AI Overview eligibility are the standard preview directives.

Why did my impressions rise while my clicks fell?

That pattern often reflects fan-out exposure rather than a ranking problem. Your page may be supporting answers across a spread of related questions, with the answer delivered on the results page. The Search Console generative AI report is the first native way to check what share of impressions came from AI surfaces.

Do I need an llms.txt file?

Not for Google. Google Search does not use llms.txt or similar AI files, and they neither help nor harm your visibility or rankings there. Other AI systems do read them, so maintaining one is reasonable if those systems matter to you. It will do nothing for your Google AI Overview citations.

How is getting cited different from ranking?

A ranking is a position for one query. A citation is a supporting link attached to one generated answer, and because Google fans a question out into several related searches, one page can be cited across questions you never targeted. Page-one visibility helps, but position and citation are separate outcomes.

Do AI Overviews appear on every health search?

No. Google shows them only when its systems judge them additive to classic Search. Published estimates for healthcare range from 51% of health queries in one analysis of over 130,000 queries to roughly 88% in a separate industry study. The studies used different query sets at different times, so treat both as estimates.

Does this apply to ChatGPT and Perplexity?

Only partly. This guide covers Google Search specifically. ChatGPT, Perplexity, Claude, and Gemini select and cite sources through different systems, and some read files that Google ignores. Verifiable authorship and directly answered questions travel well across all of them. The technical specifics here do not.

How long after fixing a snippet directive will I see a change?

It depends on recrawl timing. Google says crawling can take anywhere from several days to several months depending on how often its systems decide a page needs refreshing, and you can request a recrawl. Use the URL Inspection tool first to confirm Googlebot is receiving the corrected HTML.

Definition Bank

Entity Cards

Google AI Overviews

Search Console Generative AI Performance Report

Structured Data for AI Features

Sources

1. Google Search Central, AI Features and Your Website, updated December 10, 2025. developers.google.com
2. Google Search Central, Optimizing Your Website for Generative AI Features on Google Search, updated July 10, 2026. developers.google.com
3. Google Search Central, Robots Meta Tag, data-nosnippet, and X-Robots-Tag Specifications, accessed September 2026. developers.google.com
4. Google Search Console Help, Search Generative AI Performance Report, launched June 3, 2026, controls available worldwide August 31, 2026. support.google.com
5. WebFX, AI Overviews in Healthcare: What Our Study of 130,000+ Health Queries Reveals, September 2025. webfx.com
6. BrightEdge, Healthcare and AI Overviews: How Google Sharpened Its Approach Over Three Years, December 2025. brightedge.com
7. Aggregated AI citation analysis, The State of AI Citations 2026, drawing on Peec AI citation data, 2026. Domain-share figures are third-party estimates
8. Google Search Central, Creating Helpful, Reliable, People-First Content, accessed September 2026. developers.google.com
9. Google Search Central, Structured Data General Guidelines, accessed September 2026. developers.google.com
10. Ziptie, Google AI Overviews Source Selection, analysis of 2,400 citations, 2026. ziptie.dev
11. CXL, Google AI Overview Citation Sources, study of 100 cited pages, 2026. cxl.com