AI search and answers
AI Overviews on legal and health queries: what the data shows
Written by the Top SEO Experts Team
Reviewed by Ritik NamdevLast updated 22 min read
Quick answer
AI Overviews on legal and health queries appear often enough to matter, but published rates swing widely. Ahrefs found them on 44.1% of medical and 23.6% of legal YMYL keywords in September 2025. SE Ranking found 77.7% of its legal keywords had one a year earlier, so treat each rate as a snapshot.
Definition An AI Overview is the summary written by AI that Google can show on a search results page.
Picture a personal injury partner who opens Search Console. Impressions are up and clicks are flat. The partner decides AI Overviews ate the leads.
Maybe they did. Or a ranking slipped, or a page that used to win the click now gets answered in the snippet. The chart alone can’t say which.
What does this page do, and what does it not do?
We haven’t run our own prompt test for this page. We won’t pass other people’s numbers off as ours.
Below you will find the published evidence, why it contradicts itself, and a protocol you can run on your own queries.
How often do AI Overviews appear on legal and health searches?
It depends on the keyword set and the date. Two large studies a year apart looked at this. They came from SE Ranking and Ahrefs.
They put legal rates at 77.7% and 23.6%, and health rates between 43% and 65%. Neither describes your queries. Together they show the range.
Here are the two studies side by side.
| Study | Data collected | Keyword set | Legal | Health or medical |
|---|---|---|---|---|
| SE Ranking | 15 Sep 2024 | 300 legal and 300 health keywords, many long-tail, US desktop | 77.67% (233 of 300) | 65.33% (196 of 300) |
| Ahrefs | Sep 2025 | 146,122,391 desktop SERPs from its own database | 23.6% (legal YMYL) | 44.1% (medical YMYL); 43.0% (health industry category) |
What do the terms mean?
A long-tail keyword is a longer, more specific search phrase. A SERP is one search results page.
YMYL stands for “Your Money or Your Life” topics, where bad information can harm someone.
What was the baseline across all searches?
Ahrefs also reported a baseline. Across all 146 million SERPs, AI Overviews showed on 20.5%.
YMYL keywords triggered them 34.3% of the time. Non-YMYL keywords triggered them 17.2% of the time.
Source: What Triggers AI Overviews
View the data as a table
| Item | Value |
|---|---|
| All SERPs | 20.5% |
| YMYL keywords | 34.3% |
| Non-YMYL keywords | 17.2% |
So the direction agrees. Health and legal topics trigger Overviews more than average, but the size of the gap does not agree.
What do both studies have in common?
- Both measured desktop results.
- SE Ranking used a US location. Ahrefs does not say which country its SERPs came from.
- Both come from companies that sell SEO software.
Why do the published numbers disagree?
Because they answer different questions. A fixed list of 300 keywords per topic is not the same as a database of 146 million results. Neither one matches your own keyword list.
Five differences account for most of the gap.
| Difference | What it does to the rate |
|---|---|
| Keyword selection | SE Ranking says keyword choice shapes its results, and many of its keywords were long-tail. Ahrefs shows informational queries trigger Overviews most, so a long-tail question list skews high. |
| Date | Google decides when an Overview adds value, and that decision moves. The two studies sit 12 months apart. |
| Categories | Ahrefs uses its own internal categorization. “Legal YMYL” there is a labeled subset, not every query a law firm cares about. |
| Location | SE Ranking watched one legal query return South African information, then UK information, then no Overview at all. |
| Interest | Both firms sell tools that benefit when AI Overviews look like a big deal. That doesn’t make the data wrong, but you should read the method before the headline. |
Key takeaway Published AI Overview rates differ because the keyword sets, dates, places and categories differ. Read the method before you read the headline number.
What should you ask before trusting a new AI Overview study?
A new headline number will arrive soon. Run it through this list before it reaches a partner meeting. It’s our checklist, not a standard.
- How many keywords, and who picked them?
- What dates, and was each keyword checked once or repeatedly?
- Which device and which location?
- What counts as “legal” or “health” in the study’s own categories?
- Does the publisher sell something the finding helps?
If a study can’t answer the first three, we don’t use its number. Use rates like these to size the question, not to forecast your results.
Our view A firm whose calls come from “car accident lawyer near me” has a different exposure than one that publishes guides on statutes of limitations.
Which legal and health queries are most likely to trigger one?
Informational ones, by a wide margin. Ahrefs measured AI Overviews on 21.4% of informational keywords. It measured only 4.3% of commercial, 2.1% of transactional and 0.9% of navigational ones.
Local queries showed them 7.9% of the time.
The chart shows the same numbers. Informational searches sit far above the rest.
Ahrefs didn’t publish rates for specific legal or medical phrases. The table below is our classification of hypothetical queries. It does not show measured results.
| Example query (hypothetical) | Likely intent group | What Ahrefs measured for that group |
|---|---|---|
| what is the statute of limitations for a car accident | Informational | 21.4% |
| car accident lawyer near me | Local, commercial | 7.9% local, 4.3% commercial |
| what is a normal ALT level | Informational | 21.4% |
| dentist accepting new patients | Local | 7.9% |
| hire a personal injury attorney | Transactional | 2.1% |
The searches that fill a calendar mostly sit in the low-trigger groups. The searches people use to judge whether a practice sounds credible sit in the high-trigger group.
That’s a hypothesis about exposure, not proof of lost leads.
A sizing example with invented numbers
Suppose a firm tracks 200 informational queries and 100 hire-intent queries. We apply the all-industry intent rates. That isn’t how Ahrefs measured its legal segment, so call it a rough, hypothetical sketch.
Example Informational: 200 times 21.4% is about 43 queries that might show an Overview on a given day. Hire-intent: 100 times 2.1% is about 2. This is a hypothetical sketch.
That gap, 43 against 2, is the whole point. Exposure concentrates in the pages that explain. It barely touches the pages that book.
What should you do with that?
In our view, check where your search clicks come from. If more than half come from informational pages, build the test described below before you change anything.
If most come from local and hire-intent pages, put the effort into the map pack and service pages. Then check Overviews quarterly.
What does Google say about appearing in AI Overviews?
Google’s position is documented and short. A page must be indexed and eligible to appear in Search with a snippet. Beyond that, Google’s AI features documentation is plain (updated 10 Dec 2025, accessed 22 Oct 2026).
It says: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”
The same page adds several points worth knowing before you pay anyone for AI Overview optimization.
- Meeting the requirements guarantees nothing. In Google’s words, “Indexing and serving isn’t guaranteed.”
- AI Overviews and AI Mode may use query fan-out. That means they run several related searches across subtopics. So cited links can differ from the classic top 10.
- Sites cited in AI features count toward overall search traffic. They appear in Search Console’s Performance report under the Web search type. There’s no separate AI filter.
- Google says there is “no special schema.org structured data that you need to add,” and that structured data should match the visible text.
- The controls are
nosnippet,data-nosnippet,max-snippetandnoindex. Robots.txt directives for Googlebot control crawl access. If content still shows after you add one, Google says to check the markup in the URL Inspection tool and allow time for a recrawl.
Google also says it shows AI Overviews only where they add value beyond classic results. So most queries never get one.
Watch out Meeting Google’s requirements does not promise a spot in an Overview. Be careful with anyone who sells a markup fix or a special format as a sure way in.
What does Google say about YMYL content?
For YMYL content, the key guidance is Google’s page on helpful, reliable, people-first content, updated 5 October 2026. It says “trust is most important” among the E-E-A-T aspects. It also says E-E-A-T itself isn’t a specific ranking factor.
The page asks who wrote a piece of content, how it was produced and why it exists. For legal and health pages, we read that as a bar to clear. There’s no trick in it.
What is documented and what is theory?
Everything in this section is documented Google behavior. Claims that a schema type or summary box earns citations go beyond it, and we count those as experimental or theory.
- Documented by GoogleIndex the page, allow a snippet, and use the listed controls. No special markup is needed.
- Experimental or theoryClaims that a schema type or summary box earns citations.
Our guide to ranking in AI Overviews sorts the common claims the same way, and our schema post covers the markup question.
Want to apply this to your site?
Not sure where to start? Get a plan for your own business.
What happened when AI Overviews got health answers wrong?
Google removed some health Overviews after a newspaper showed errors. The Guardian reported on 2 January 2026 that AI Overviews gave misleading information on some health queries.
One example was the normal range for liver blood tests. The numbers ignored age, sex, ethnicity and nationality.
The Guardian followed up, and TechCrunch summarized that follow-up. By 11 January, Overviews had been removed for two queries. They were “what is the normal range for liver blood tests” and “what is the normal range for liver function tests.”
What did Google and The Guardian say next?
Per The Guardian, variations such as “lft reference range” could still lead to AI-generated summaries. TechCrunch’s writer tried them and saw no AI Overviews.
Google told The Guardian it doesn’t comment on individual removals. It said it works on broad improvements. It also said clinicians reviewed the flagged queries and found that “in many instances, the information was not inaccurate and was also supported by high quality websites.”
What can a practice take from this?
What a practice can take from this is our reading. It is not Google’s reading.
- Removals happen query by query and without announcement. An Overview you see today may be gone next month. One that’s gone may come back.
- Variants of a removed query can keep their Overviews. So testing a single phrasing proves little.
- Numbers without context were the failure that got flagged. Don’t publish a lab range, dose or deadline on your own pages without the context that changes it. Examples are age, sex, state or filing type.
Which sources do AI Overviews cite on these queries?
In the only public source breakdown we found for these categories, large institutions and publishers dominate. SE Ranking’s September 2024 set showed Mayo Clinic with 107 links, WebMD 91, Healthline 77 and Medical News Today 74 for health. For legal, NYCourts.gov had 114, YouTube 48 and FindLaw 46.
Does ranking well still matter?
SE Ranking also measured overlap with organic results. On average, 7.13 links in a health Overview also appeared in the top 20 organic results for the same query. For a legal Overview the average was 6.49.
Ranking well in classic search still looks like the surest route in. Overlap isn’t causation, though.
Should you trust the directory claim?
A June 2026 report from Haute Lawyer Network and 5W comes from a PR and generative engine optimization agency. It says directories supply most of the citations in ChatGPT, Claude, Perplexity and Google AI Mode. The directories it names are Chambers, Legal 500, Super Lawyers, Best Lawyers, Martindale, Avvo and Justia.
Watch out We wouldn’t rely on that report. It doesn’t state its query count and gives no run dates. Its authors also sell a legal GEO audit.
Our hypothesis is that on YMYL queries a practice competes with courts, hospitals and national publishers. So the opening is detail they can’t match.
Examples are your state’s statute, your county’s filing process and your clinic’s real intake steps. We have no data showing this earns citations.
What should a law firm or clinic do about AI Overviews?
Do the fundamentals first, because Google’s documentation says nothing else is required. This is the order we use on YMYL sites. It’s our method, not a Google checklist.
- Confirm indexing and snippet eligibility: run your main informational pages through URL Inspection in Search Console, and check that no
nosnippetrule is blocking them. Our free Website SEO Checker covers basic crawl and indexing signals. - Sort target queries by intent. Protect the pages that book consultations and appointments first. Informational pages come second, because that’s where Overviews cluster.
- Show who wrote and reviewed each page: name the attorney or clinician, add the date and state the jurisdiction. That answers Google’s who, how and why questions, and our YMYL editorial standards post shows how we apply it.
- Add the detail an Overview can’t. Use state-specific deadlines, local court steps, or the clinical context behind a range or dose.
- Run a compliance read before publishing. New Hampshire’s version of Rule 7.1 says “A lawyer shall not make a false or misleading communication about the lawyer or the lawyer’s services.” Your state’s wording differs, so check it, and see our post on attorney advertising rules and SEO.
- Measure in Search Console. Watch impressions and clicks by query and page for your informational group before and after any change.
What are our four rules of thumb?
When a partner asks what to do about all this, we fall back on four rules.
- If the answer depends on jurisdiction, write the state page. Skip chasing the Overview.
- If an Overview shows wrong information about your own practice, fix the source pages first. Then document the date and wording.
- If a page makes a medical or legal claim you wouldn’t say to a client in person, it doesn’t get published. Traffic potential doesn’t change that.
- If clicks fall but calls hold steady, don’t rewrite anything yet. Calls are the number that pays.
How do you read a falling click-through rate?
Take a hypothetical informational page group. Over one 90-day window it earns 40,000 impressions and 1,200 clicks. That is a 3.0% click-through rate.
The next 90 days it earns 42,000 impressions and 1,000 clicks, or 2.38%. Clicks fell about 17% (200 fewer, divided by 1,200).
| Hypothetical 90-day window | Impressions | Clicks | Click-through rate |
|---|---|---|---|
| First window | 40,000 | 1,200 | 3.0% |
| Next window | 42,000 | 1,000 | 2.38% |
That pattern fits an Overview taking some clicks. It equally fits a ranking slip, a seasonal change or new competitors. A falling rate beside rising impressions can also just mean wider visibility (why click-through rate falls as visibility grows).
Only the test below separates them. For the wider picture, see zero-click searches.
Where can you read more?
Most of this is ordinary SEO for a regulated profession. Firms can read our law firm lead generation post and legal industry page.
Dental and medical offices can start with dental patient acquisition. Our AI search optimization service covers where AI search fits in a wider program.
How do you test AI Overviews for your own practice?
Run a fixed prompt set under fixed conditions, on three separate days, and log every result. The goal is a dated baseline for your own queries, not a universal rate. No published standard says how many runs are enough.
Why three days?
SparkToro and Gumshoe.ai reported in January 2026 that ChatGPT and Google’s AI returned the same brand list less than 1% of the time. That was across repeated runs of the same prompt.
The study covered brand recommendation lists, not legal or health Overviews. It’s still a fair warning against trusting one look.
What is the protocol?
Here is the protocol we’d use for any practice.
- Build 40 queries: 10 informational, 10 local, 10 hire-intent, 10 branded or comparison.
- Use a logged-out browser profile and a fixed US location, and note desktop or mobile.
- Run the set on three different days, ideally across a week.
- For each query, record whether an Overview appeared, which domains it cited, and whether your site was among them.
- Save a screenshot with the date visible.
- Cross-check against Search Console impressions for the same queries.
- Build the query setMix informational, local, hire-intent and branded queries.
- Fix the conditionsUse a logged-out profile, one US location and one device.
- Run it on three daysSpread the runs across a week.
- Log each resultNote the Overview, the cited domains and your own site.
- Save dated screenshotsKeep the date visible in each image.
- Check Search ConsoleCompare against impressions for the same queries.
A log can be as simple as this template, filled with your own queries.
date | device | location | query | intent | overview? | cited domains | us cited? | screenshot
2026-10-26 | desktop | Austin TX | (query) | informational | yes | (list) | no | file nameHow do you read the result?
Read the result by group, not by query. Our labeling is a convention and not a standard.
| Result across the 3 days | Our label |
|---|---|
| An Overview on all 3 days | Stable |
| An Overview on 1 or 2 days | Unstable |
| No Overview on any day | Absent |
Unstable queries tell you the most about how little a single check proves.
If informational queries are mostly stable and hire-intent queries mostly absent, you match the pattern Ahrefs measured. If not, trust your own data. Repeat quarterly.
Quick check Can you show a dated log of your own queries on three separate days? If not, you are working from someone else’s numbers.
Our guide to measuring AI visibility covers the wider measurement problem. What GEO tools can’t measure explains why even paid trackers have limits.
What does the data not show?
Quite a bit. As of 22 October 2026, this is what we couldn’t find evidence for.
- Lead impact. We found no public data tying an AI Overview citation to calls, consultations or booked appointments for a law firm or clinic.
- Format effects. No study we opened shows that FAQ blocks, summary boxes or any schema type increase citations on legal or health queries. Google’s documentation says no special markup is needed.
- Accuracy comparisons. SE Ranking describes a manual review of specific examples alongside a broader quality assessment. We do not read it as a systematic accuracy evaluation, so it doesn’t show whether legal or health answers are more accurate.
- Removal scale. We know of two named queries removed in January 2026. The total is unknown.
- Directory dominance. The claim comes from a vendor report covering other engines, with no reproducible method.
- Practice-area and state differences. None of the studies we opened break rates down by practice area or US state.
If a vendor tells you otherwise on any of these, ask for the method and the dates. Our post on LLM SEO myths lists more claims worth challenging. For the definition of the term itself, see our AI Overviews glossary entry.
What this means for you
The rates in this post are snapshots, so use them to size the question. Your own queries and your own dated log are what count.
If most of your clicks come from pages that explain, expect more exposure to Overviews. If most come from pages that book, expect less. Start by finding out which group you are in.
Then do the fundamentals. Make sure your pages are indexed and can show a snippet. Name the attorney or clinician who reviewed each page, and keep every legal and health claim in line with your state’s rules.
Watch out This page is general information, not legal or medical advice. Rules differ by state and change over time, so check the current rules for your own situation.
Frequently asked questions
Do AI Overviews appear on law firm searches?
Sometimes. Ahrefs measured them on 23.6% of legal YMYL keywords in September 2025. SE Ranking saw them on 77.67% of a 300-keyword legal set in September 2024.
Keyword sets and dates explain the gap. Informational legal questions trigger them far more often than hire-intent searches.
Does Google show AI Overviews for every health question?
No. Ahrefs found them on 44.1% of medical YMYL keywords in September 2025, so most had none. Google also removed Overviews for at least two liver test queries in January 2026 after a Guardian investigation.
The Guardian reported that close variants of those queries could still produce summaries. TechCrunch’s writer tried them and saw no AI Overviews.
Can I stop my pages from appearing in AI Overviews?
Google lists nosnippet, data-nosnippet, max-snippet and noindex as the controls. They work for Search as a whole, so a nosnippet page loses its ordinary snippet too.
If the content still appears afterward, Google suggests rechecking with URL Inspection. It also suggests allowing time for a recrawl. Your compliance counsel may reasonably choose differently.
Do I need special schema or an llms.txt file to be cited?
No. Google’s AI features page, accessed 22 October 2026, says no additional requirements apply. You needn’t create new machine-readable files, AI text files or special schema.org markup.
The page must be indexed and eligible to show a snippet. Treat any pitch for a markup fix to win citations as theory.
Is it safe to use AI to write legal or health pages?
Only with accountable human review. A licensed attorney or clinician should check each claim, jurisdiction and date before anything goes live. That’s our view, not a Google rule.
Attorney advertising rules, such as Rule 7.1 on misleading communications, apply to web content however it was drafted.
Where do AI Overview clicks show in Search Console?
Google says sites cited in AI features are counted in the Performance report under the Web search type. There is no separate AI filter.
We haven’t found a reliable way to isolate those clicks. Track query and page trends over matching periods. Be skeptical of any tool that claims exact attribution.
Next step
Want a baseline before you spend anything on AI search? Start with the free Website SEO Checker for crawl and indexing basics. Then Get Your Organic Growth Plan.
Sources
- Google Search Central, AI features and your website, last updated 10 Dec 2025 on the page, accessed 22 Oct 2026.
- Google Search Central, Creating helpful, reliable, people-first content, last updated 5 Oct 2026 on the page, accessed 22 Oct 2026.
- Ahrefs, What Triggers AI Overviews? 86 Factors and 146 Million SERPs Analyzed, accessed 22 Oct 2026.
- SE Ranking, AI Overviews Research: Analyzing Google’s approach to YMYL topics, accessed 22 Oct 2026.
- TechCrunch, Google removes AI Overviews for certain medical queries, accessed 22 Oct 2026.
- Haute Lawyer Network and 5W, The 2026 Legal AI Visibility Report, accessed 22 Oct 2026.
- Search Engine Journal, AI recommendations change with nearly every query: SparkToro, accessed 22 Oct 2026.
- State of New Hampshire Judicial Branch, Rule 7.1 Communications Concerning a Lawyer’s Services, accessed 22 Oct 2026.
Want to apply this to your site?
Not sure where to start? Get a plan for your own business.
