General
Why Clinic Website Traffic May Fall as AI Answers Expand in 2026
A clinic can remain visible in search while receiving fewer website visits. An AI summary may answer a question on the results page, but that is one possible explanation to test alongside changes in demand, rankings, competition and tracking.
Someone may read a summary, follow a citation, open a map listing or search again. Each action can lead to a different kind of visit or inquiry.
The Korean article used Opensurvey's 54.5% recent ChatGPT-use figure to describe the changing environment. That survey measure does not establish that every Korean patient's first search has moved to AI, or identify the cause of one clinic's traffic decline.
This guide reviews the research and the checks a clinic can make. The Korean original was published April 22, 2026; this English edition checks sources and feature information as of October 31, 2026.
1. International studies show different click patterns when AI summaries appear
Pew Research Center's July 2025 analysis
Pew examined browsing by 900 US adults during March 2025, with 68,879 unique Google searches. It reconstructed results in April, so the classification reflects the results collected then, rather than a guaranteed replay of what each person originally saw.
- Approximately 18% of searches in the study produced an AI summary.
- Visits to results without a summary led to a traditional result-link click about 15% of the time.
- For results with a summary, the corresponding figure was 8%.
- A link inside the summary was clicked in 1% of visits to pages with a summary.
- A browsing session ended after 26% of summary-result visits, compared with 16% without one. This does not establish that every person ended because the answer satisfied them.
These are observed behaviors in a US sample, not a randomized causal experiment or a Korean clinic conversion forecast.
Seer's historical CTR study
Seer's September 2025 analysis, published November 4, covered informational and educational queries. It reported the following historical figures:
- Organic CTR for its AI-Overview query group moved from 1.76% to 0.61%. Those endpoints imply approximately a 65.3% decline, although the source prints 61%; the source's percentage and endpoints do not reconcile.
- Paid CTR moved from 19.70% to 6.34%, approximately 68% lower.
- Organic CTR for the non-Overview group moved from 2.72% to 1.62%, approximately 40.4% lower; the source reports roughly 41%.
The figures describe that dataset, not every clinic. Seer's April 24, 2026 follow-up reported that the earlier downward trend was leveling off or reversing in newer data. Do not turn the older series into a prediction of uninterrupted decline.
Check what each zero-click figure actually counts
The original article cited a rise from 56% in May 2024 to 69% in May 2025. Similarweb's own public explanation identifies these as news-related Google searches, not all searches or clinic searches. A news-publisher statistic cannot establish that seven out of every ten prospective clinic patients never visit a website.
2. Korean survey and product data need their own context
Opensurvey's first-half 2026 report
The report compared March and December 2025 samples of 1,000 respondents each, aged from their teens through their fifties. The December fieldwork was December 9–10. These recent usage-experience rates can overlap:
| Channel | Reported usage experience | Reported comparison |
|---|---|---|
| Naver | 81.6% | Lower than the earlier comparison |
| YouTube | 72.3% | Lower |
| ChatGPT | 54.5% | Approximately 15 percentage points higher |
| Gemini | 28.9% | Approximately three times the earlier level |
They are not mutually exclusive market shares, a census of all Koreans or a patient-only survey. The original also quoted earlier 10% usage and 83.3% paid-subscriber-share reports. Those use different populations and dates; they should not be joined into one consistent national time series without checking the original methods.
Distinguish Naver's products, their users and the measured industries
The original discussed the reported April 2026 closure of Cue and a move toward integrated AI search. Product names and availability can change; this does not prove that Naver has disappeared as a patient channel or that every clinic is now summarized in the same way.
A directly checked Naver announcement dated August 21, 2025 described Place AI Briefing for restaurants, cafes and other places. Its reported before-and-after measures included:
- 10.4% longer time spent in the measured Place search area.
- 27.4% higher click-through.
- 137% more clicks on the “more” tab.
These are the provider's reported measures for participating businesses, not a controlled study showing the same improvement for hospitals. The announcement discussed future industry expansion. Inspect the actual current result before assuming a clinic is eligible, ranked by a particular formula or shown in a fixed three-to-four-clinic summary.
3. Health-query visibility and clinic recommendations are different measures
Ahrefs' September 2025 desktop results
Ahrefs' November 10 publication analyzed 146 million desktop search results from September 2025:
- AI Overviews appeared for 44.1% of queries in its medical-YMYL category.
- The overall share across its dataset was 21%, so the medical-category figure was more than twice that level.
This measures whether a summary appeared. It does not show that four or five out of ten Korean patients read only the summary, or that every medical query behaves like a local appointment search.
Which sources appeared in a health-information study?
SE Ranking's December 2025 snapshot examined 50,807 German-language health queries from Berlin. It classified 34.45% of citations as coming from sources with the editorial or clinical safeguards defined by the researchers.
- Academic journals, domestic governmental health sources and international governmental sources accounted for 0.48%, 0.39% and 0.35%, respectively, in its classification.
- YouTube accounted for 4.43% of citations. A platform share is not evidence that a Korean clinic must publish physician interviews to be eligible.
Health information, video discovery and a local clinic shortlist should not be treated as one identical outcome. A relevant, clinically responsible video may help an audience, but the study does not justify replacing sound clinical explanations with a video quota.
Citation does not certify accuracy or authorship
Originality.ai classified 10.4% of cited documents as likely AI-generated in its YMYL study, using its automated detector. That is a model-based classification, not proof of who wrote every document, and it does not mean that every flagged statement is false.
A cited page still needs evidence and appropriate clinical responsibility. The same applies to the clinic's own material.
4. Four patterns to investigate in the clinic's own records
The following are working scenarios, not newly verified observations from BlinkAd clients.
1. Impressions hold up while clicks fall. Compare query groups, average position, country, device, dates and result features. AI summaries may contribute, but rank changes, demand and measurement issues can produce a similar pattern.
2. The same publishing effort brings fewer inquiries. The original's ten-post example—3,000 visits and thirty inquiries falling to 1,500 visits—is an illustrative calculation without a supplied clinic dataset. Track actual outcomes rather than treating it as an old industry formula.
3. Competitors appear in relevant AI answers. Record repeated responses and cited sources. Being named is different from gaining patients, and being absent from one response is different from losing all new inquiries.
4. Local discovery includes AI summaries. Check what Naver and other services actually show for the clinic's questions. Reviews, recency and mentions are not a verified universal three-factor formula that determines every summary.
5. Extend useful SEO work into an answer-focused review
AEO adds attention to how a question is answered and how the clinic is represented. It does not replace accessible pages, useful explanations or reliable search fundamentals.
We use the five areas below as a review framework. The original article did not supply the underlying clinic sample needed to establish them as findings from a new BlinkAd study.
| Area | What to review | What the evidence does not establish |
|---|---|---|
| Identity and responsibility | Accurate institution, physician and authorship information. | No single universal entity signal guarantees recommendation. |
| Supported information | Appropriate clinical evidence and properly defined actual records. | The GEO paper's experimental improvements up to 40% are not a minimum clinic citation uplift. |
| Clear answers | Direct explanations with necessary context and useful FAQs. | FAQ or HowTo markup is not mandatory for AI inclusion, and Google no longer displays FAQ rich results. |
| Current facts | Substantive revisions and accurate dates when information changes. | Ahrefs' 76.4% recent-date observation uses 564 detectable-date URLs, not every citation in a separate 17-million study. |
| Search access and visibility | Crawlability, indexing, relevant queries and the actual results. | A top-ten organic position is not a universal prerequisite for an AI citation. |
Originality.ai's ranking study illustrates why denominators matter: its top-ten percentage concerns only citations already overlapping the top hundred organic results. It excludes the separate group outside that range. A checklist is useful for finding gaps, not for declaring a clinic ineligible whenever one area is weak.
6. Three checks to begin with
1. Compare six months of Search Console impressions and clicks. Segment relevant queries and pages, check positions and dates, and confirm tracking. Stable impressions with fewer clicks create a question to investigate, rather than proving that AI absorbed the traffic.
2. Check three relevant patient questions in available AI search services. Choose questions about the clinic's actual services and location, record the conditions and repeat the checks. Distinguish mentions, linked citations, visits and confirmed inquiries.
3. Inspect the actual local search experience. Check the clinic's Naver Place and other relevant results, accurate business information and available summary features. Do not assume a product rollout or a ranking factor without evidence.
Preserve useful existing material and improve what patients cannot currently find or understand. Then compare results over an appropriate period, accounting for other changes.
Contact BlinkAd about a clinic search and content review. We can discuss the Search Console patterns, observed AI answers and relevant local information, with clinical and advertising questions directed to the appropriate expertise. Citation, traffic and patient growth remain separate outcomes to measure.
Sources and reading notes
- Pew's July 22, 2025 study; March US browsing and April result reconstruction, not a clinic experiment.
- Seer's historical study; the printed 61% and 1.76-to-0.61 endpoints conflict arithmetically.
- Similarweb's own explanation of the 56-to-69% news-query series; not an all-query rate.
- Seer's April 24, 2026 follow-up; newer data qualified the earlier continuous-decline prediction.
- Opensurvey's public findings and sample and dates.
- Naver's August 21, 2025 announcement; measured Place businesses and restaurant/cafe context.
- The original Cue closure and later AI Tab accounts are historical product background; a current clinic's availability and presentation need direct confirmation.
- SE Ranking's primary German health-source study; category and citation classifications are specific to its snapshot.
- Originality.ai's AI-classification study; automated estimates with an unclassifiable group, not verified authorship.
- The earlier 10% use and 83.3% subscriber-share accounts cited by the Korean article use distinct populations and are not joined to Opensurvey as one time series.
- Ahrefs' 146-million desktop-result analysis, September 2025 data, November publication.
- Originality.ai's ranking analysis; its top-ten finding is conditional on the top-hundred overlap subset.
- The GEO paper and Ahrefs' top-cited-page analysis; experimental improvement and recency observations are not clinic guarantees.
BlinkAd is a Google and AI marketing agency helping businesses prepare for international customers' searches, visits and inquiries.