Will AI Overviews Reduce Patient Traffic? What the Data Shows

Partly, and only to specific pages. AI Overviews now appear on nearly every clinical search, so condition, symptom, and treatment content is losing clicks. The searches that book appointments, meaning provider and “near me” queries, carry far less coverage. Which pages you rank for decides your exposure.

TL;DR

• Query type, not specialty, determines whether AI Overviews affect your traffic. Published healthcare figures range from 17 to 89 percent because each study samples a different set of keywords.
• Clinical content is near-saturated. Treatment queries returned an AI Overview 100 percent of the time in December 2025, symptom queries 93 percent.
• Provider and “near me” searches were tested and pulled back. Readings range from 0 to about 12 percent, an order of magnitude below clinical content.
• The most-quoted statistic is misreported. A fall from 1.76 to 0.61 percent is a 65 percent decline, not 61, and the same study found a 41 percent decline on queries with no AI Overview at all.
• Click-through rate can fall while clicks hold steady, because rising impressions grow the denominator.
• Early 2026 data shows click-through rates recovering from their December 2025 floor, which makes the collapse narrative already dated.

One liner: AI Overviews are taking clicks from the pages that answer medical questions, not from the searches that book appointments, and the size of the effect has been overstated by the numbers in widest circulation.

Key Numbers

Whether AI Overviews Cost You Patients Depends on One Variable

In short: AI Overviews reduce clicks on some healthcare searches and none at all on others. The variable that decides which is query type, not specialty, and not industry.

Published figures for how often AI Overviews appear on healthcare searches range from roughly 17 percent to 89 percent [1][2]. Both numbers are accurate. They measure different things. BrightEdge tracked clinical informational queries, the symptom explainers and treatment comparisons, and found near-saturation [1]. SE Ranking sampled healthcare queries broadly, including local provider searches that show almost no AI Overview presence at all, and landed at 17 percent [2].

That spread is the finding. “Healthcare” isn’t a keyword population. It’s a label sitting on top of several populations that behave nothing alike. A practice publishing condition explainers and a practice ranking mostly for “orthodontist near me” carry completely different exposure, and an industry average tells neither one anything it can act on.

So the useful question isn’t how much traffic you’ll lose. It’s narrower. Which of your pages answer questions Google now answers on the results page itself?

For most practices, that turns out to be a specific and identifiable slice of the site, and the rest is untouched. What follows is what the available data can tell you about which slice is yours, and where it goes quiet.

Four Studies, Four Different Samples, One Set of Caveats

In short: Four datasets carry most of what’s known about AI Overviews and healthcare search. Each measures something different, and none of them measures all of it.

Click-through rate, throughout this article, means clicks divided by impressions. It’s the share of people who saw your listing and chose it, which matters for section five.

Two of these sources are independent measurement. Two come from companies selling AI search visibility tools, which doesn’t make them wrong, though it does mean the finding and the product point the same direction. All four are used here, and which is which stays labeled.

Worth stating plainly: MIT’s health figures rest on 808 queries out of 11,372 sampled [4]. That’s the smallest number in this article and it appears here rather than in a footnote, because a percentage without its sample size is an opinion wearing a decimal point.

One note on age. Pew’s behavioral data was collected in March 2025 [3]. It’s the best available measurement of what people do when a summary appears. It isn’t a picture of today.

Finding 1: Your Condition and Treatment Pages Sit Under an AI Overview Almost Every Time

In short: Clinical informational content is where AI Overview coverage is effectively total. Treatment and symptom queries now return a summary nearly every time, and coverage has climbed every year since 2023.

BrightEdge has taken December snapshots of the same healthcare keyword set three years running. Coverage across that set moved from 59 percent to 89 percent over two years, and conditions and symptoms content went from 82 percent in May 2024 to 93 percent by December 2025 [1]. Within that keyword set, the breakdown by clinical sub-topic:

Query length sharpens the picture. A WebFX analysis of 130,070 US health queries collected in July 2025 found that health searches of seven words or more returned an AI Overview roughly three times in four, against about 36 percent for one- and two-word searches [6]. The longer and more specific the question, the likelier Google answers it in place. Hold onto that one, because long-tail educational content is what most practice blogs are built from.

A second sample points the same direction. The MIT team found health queries returning AI results rose from 51 percent in 2024 to 58 percent in 2025, running identical queries in both years [4]. A lower number than BrightEdge’s, drawn from a different keyword population, moving the same way.

One thing this section doesn’t claim: that any of this cost a practice a patient. It says the summary appeared. Whether people stopped clicking is section five.

Finding 2: The Searches That Book Appointments Were Largely Left Alone

In short: Google put AI Overviews on “near me” healthcare searches, then took most of them back off. Provider-intent queries carry a fraction of the coverage clinical content does, though the exact fraction is disputed.

This is where the reporting gets messy, and the mess is worth walking through rather than around.

BrightEdge published two tracking series covering local healthcare provider queries over overlapping periods, and they don’t agree. One reports these queries at 100 percent AI Overview presence in December 2023, falling to 14 percent by December 2024 and 0 percent by December 2025. The other, tracking searches like “doctor near me” and “urgent care near me,” reports 0 percent in May 2024, rising to 4 percent that December, 14 percent by May 2025, and 11 percent by December 2025 [1]. Same vendor, opposite trajectories.

So no single figure here is defensible. The range is:

Four readings, none above 15 percent. Set that against treatment queries at 100 percent and symptom queries at 93 percent from the last section, and the conclusion worth drawing isn’t a number. It’s an order of magnitude. The pages answering medical questions are saturated. The searches producing appointments are not.

Between those two groups sit service and department pages, which mix informational and navigational intent. Published measurement on this middle tier is thin. The reasonable expectation is a milder decline than clinical content shows, and that’s inference rather than a finding.

The mechanism looks straightforward. These queries belong to maps and the local pack, and synthesis adds little when someone wants a phone number and directions.

One caution. “Largely left alone” describes now. Google tested this category once, reversed itself, and the series above suggest it’s still adjusting. Treating the current state as settled would be a mistake, which is the argument for building AI search optimization for healthcare before the picture shifts rather than after.

Finding 3: Why a Falling Click Rate Doesn’t Mean Fewer Patients

In short: The most-quoted statistic in this debate has three problems. The arithmetic is wrong, the same study found similar declines where AI Overviews were absent, and click-through rate can fall while clicks hold steady.

The figure you’ll see everywhere: organic click-through rate fell 61 percent on queries with AI Overviews, from 1.76 percent to 0.61 percent, in Seer Interactive’s analysis of 3,119 informational queries between June 2024 and September 2025 [5].

Start with the arithmetic. A drop from 1.76 to 0.61 is a decline of about 65 percent, not 61. Small gap, and it doesn’t change the story. The mislabeled figure is the part that travels, though, and several pages now print the correct percentage in a table and the wrong one in the sentence above it.

The error also multiplies. At least one page ranking for this topic publishes a table applying that same decline to treatment, pain, and symptom queries separately. Seer measured an aggregate across 3,119 informational queries spanning 42 organizations in every vertical. It never produced a healthcare-specific click-through figure, and certainly not one broken out by clinical category [5].

The second problem is bigger. That same study also measured queries where no AI Overview appeared. Click-through rate on those fell 41 percent over the same period, from 2.72 percent to 1.62 percent [5]. Click rates dropped sharply in the absence of the thing being blamed. AI Overviews are part of what’s happening to search behavior rather than the whole of it, and attributing the full decline to them overreaches the evidence.

Third, click-through rate is a ratio: clicks divided by impressions. Seer’s own reading of a later decline attributes it largely to a surge in impressions rather than a collapse in clicks [10]. A page that starts surfacing on searches it never appeared for gains impressions, and the ratio falls even when the click count holds or grows. Some practices watching CTR fall in Search Console are watching their denominator expand.

None of this makes the effect imaginary. Pew found people click a traditional result 8 percent of the time when a summary appears, against 15 percent when it doesn’t [3], and Ahrefs measured position-one impact deepening from 34.5 percent to 58 percent between April 2025 and February 2026 [8]. The direction is consistent across methods. The magnitude and the attribution are what’s soft.

One figure here points somewhere useful. Brands cited inside an AI Overview earn roughly 120 percent more clicks per impression than uncited brands on the same results page [10]. Being in the summary is worth about twice being under it.

Three Traffic Patterns, and What Each One Is Telling You

In short: Search Console can separate an AI Overview effect from a ranking problem in about ten minutes. The signature is impressions holding steady while clicks fall and position doesn’t.

A note first. Everything above this point came from a named study with a stated sample. What follows doesn’t. It’s reasoning drawn from those findings, and it’s worth labeling as such rather than letting it borrow their authority.

The healthcare-specific step is segmentation. Split your Search Console queries into two groups: condition, symptom, and treatment language on one side, “near me” and provider language on the other. Sections three and four predict those lines should be moving differently, with the clinical group absorbing nearly all of the decline. If your data shows that split, you have a content-tier problem, and the response is generative engine optimization for healthcare, which means structuring pages so an AI summary cites them instead of replacing them. If both groups fell together, something else is happening and AI Overviews aren’t the place to look first.

Use a year-over-year window rather than 28 days. Short comparisons catch seasonality and indexing noise, and they’ll tell you almost anything you want to hear.

One last check before drawing conclusions. Pull consultation requests and calls over the same period. A decline concentrated in clinical pages can leave bookings untouched, and bookings are the number the practice actually runs on.

Caveats: Five Reasons to Hold These Numbers Loosely

In short: Every figure above comes with limits. Two sources sell the solution, the key behavioral study can’t prove causation, prevalence swings month to month, and the decline may already be reversing.

1. Association isn’t causation. Pew’s data shows summaries and lower clicks occurring together. It doesn’t establish that summaries caused the drop. Google called the methodology flawed and the queryset unrepresentative of search traffic, and said it had not observed significant declines in aggregate web traffic [3].

2. Two of the four main sources sell the solution. BrightEdge and Seer both market AI search visibility products. That doesn’t make their data wrong. It does mean the measurement and the product point the same direction.

3. Large studies can have thin slices. MIT ran 24,000 queries. Of those, 808 were health queries [4]. The study is substantial. The portion applying to a medical practice is smaller than the headline number suggests.

4. Prevalence swings. Semrush tracked AI Overview presence at 6.49 percent of queries in January 2025, 24.61 percent by July, then 15.69 percent by November [11]. Google expands, evaluates, and pulls back on a monthly cycle. Any single reading is a snapshot of a moving system, which is also the likeliest explanation for the contradictory vendor series in section four.

5. The decline may already be reversing. Seer’s 2026 update shows organic click-through rate on AI Overview queries rising from a floor near 1.3 percent in December 2025 to roughly 2.4 percent by February 2026, reversing an eighteen-month slide [10]. Pages published as recently as April 2026 still describe the decline as structural rather than cyclical and report no signs of recovery. That characterization predates the data above. Two months isn’t a trend either. It’s enough to say the collapse narrative is already dated.

What to Measure Instead of Total Sessions

In short: Total sessions stopped being a useful headline metric the moment impressions and clicks decoupled. Four narrower numbers track patient acquisition better, and effort should follow the query tiers rather than spread evenly.

Clinical content gets optimized for citation. Provider-intent pages keep the local investment they already have. Four numbers worth watching in place of the session count:

• Consultation requests and calls, split by page tier
• Performance on provider-intent queries, tracked apart from everything else
• Citation presence in AI Overviews for your priority terms
• Branded search volume, which carries effectively no AI Overview exposure and often moves when visibility shifts even as clicks don’t

Earning a citation is mechanical rather than mysterious: structured answers near the top of the page, accurate schema markup, named clinical authors with stated credentials, and the E-E-A-T signals that follow from both.

One thing not to do. Clinical content that lost clicks is frequently the same content earning the citation that section five valued at roughly double the uncited alternative. Deleting or consolidating those pages to make a session chart look better would be the worst available response to this evidence.

So, will AI Overviews cost your practice patients? For the pages answering medical questions, they’re already costing you clicks. For the searches that book appointments, the evidence says not yet, and Google appears to have decided that deliberately. That’s a narrower problem than the headlines describe, and a more solvable one.

We sell this work, so weigh the recommendation accordingly. The diagnostic in section six costs nothing and doesn’t require us. If the split shows up in your data and you’d rather not run the response in-house, we can help your pages earn citations.

Frequently Asked Questions

Will AI Overviews reduce patient traffic to my practice website?

For clinical informational pages, yes, measurably. For the provider and “near me” searches that generate appointments, current data shows little to no AI Overview presence.

Which healthcare pages are most affected by AI Overviews?

Condition, symptom, and treatment pages. BrightEdge measured AI Overview presence at 100 percent for treatment queries and 93 percent for symptom queries in December 2025.

Do AI Overviews appear on “near me” searches for doctors?

Rarely. Published readings for local provider queries range from 0 to about 12 percent, an order of magnitude below clinical content. Google appears to route these queries to the local pack instead.

My rankings held steady but traffic dropped. What happened?

That pattern is consistent with AI Overview interception on informational queries. Check whether impressions held or rose while clicks fell and average position stayed flat.

Is the 61 percent click-through rate decline accurate?

The label is wrong. Seer Interactive measured organic click-through rate falling from 1.76 percent to 0.61 percent, which is a decline of about 65 percent, not 61.

What percentage of healthcare searches show an AI Overview?

There is no single figure. Published numbers range from about 17 percent to 89 percent because each study samples a different keyword population. Clinical queries sit near the top of that range and provider queries near the bottom.

Should I delete clinical content that lost traffic?

No. Pages that lost clicks are often the same pages earning AI Overview citations, and cited brands see roughly 120 percent more clicks per impression than uncited ones.

How do I get my practice cited inside an AI Overview?

Structure answers near the top of the page, apply accurate schema markup, name clinical authors with their credentials, and strengthen E-E-A-T signals.

Is the traffic decline from AI Overviews permanent?

Not established. Seer’s 2026 data shows click-through rates on AI Overview queries recovering from a December 2025 floor, and Google has expanded and withdrawn AI Overview coverage repeatedly since 2023. Descriptions of the decline as structural and irreversible are not supported by the most recent measurements.

Definition Bank

Entity Cards

Pew Research Center

Aral, Li and Zuo (MIT)

BrightEdge

Seer Interactive

Sources