The Therapist AI Visibility Index.

Source: KKV Labs, Google AI Mode answers captured October 2, 2026 and October 4, 2026 in 60 cities.

On WordPress sites we also read the page builder. As a share of every site counted: Gutenberg blocks 32%, Elementor 12%, Divi 7%, Beaver Builder 3% and WPBakery 2%. A site can use more than one.

A website vendor we could recognise sits on 8% of the sites, and Brighter Vision is the one we saw most, on 18. Most practices run a site with no vendor name on it at all.

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

How the homepage takes the first message

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

Google Analytics 4 runs on 44% of the named practices and 35% of everyone else, and a Google Ads tag on 17% against 8%. Google Tag Manager sits level, 29% against 29%. None of these tags does anything for an AI answer by itself.

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

A video is embedded on 10% of the homepages, 15% of the named practices and 6% of everyone else.

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

Blue Cross Blue Shield is the plan written most often, on 14% of the homepages. An insurance or fees page is linked from 32%.

Fees before the first call

"How much is therapy without insurance" is one of the ten client questions we track for this field, yet 2% of the homepages print a fee per session.

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

A license number is printed on 6% of the homepages. A crisis line or 988 is shown on 10%.

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

Named practices are the 117 crawled sites Google AI Mode listed in their own city's answer. Everyone else is the other 133. Only rows with at least 5 sites behind them are shown. Most of the gaps are signs of upkeep, not of what the practice offers.

The checks that read the other way round

32 checks sat within two points in both groups, among them a contact or appointment page linked; a blog or articles section linked; a services or specialties page linked; EMDR listed on the homepage. On October 5, 2026 they did not mark out the practices the answer named.

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

34% of the homepages carry only the website record a theme ships, with no business in it: 41% of the named practices and 28% of everyone else. A Person record, the markup that names the therapist, sits on 17% of the named sites and 7% of the rest.

AI crawlers are almost never shut out. Of the 242 robots.txt files we read, the most blocked AI crawler is named in a blocking rule on 2% of them. 73% to 75% of the files keep the AI crawlers out of some pages but not the whole site.

24% of the sites serve an llms.txt file, 59 sites. 29 of the 30 Wix sites we read have one, so most of these files are written by the website builder rather than by the practice.

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

What the homepage links before anyone gets in touch

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

Anxiety is named on 84% of the homepages. Telehealth is stated on 68% and in person sessions on 24%.

The lines that read the other way: written more often by everyone else

Source: KKV Labs crawl of 250 therapy practice homepages, of 275 opened, October 5, 2026.

Source: KKV Labs screening and Google AI Mode answers captured October 2, 2026 and October 4, 2026. 60 cities, 340 names. Download the city data (CSV). What we do for therapists. How we measure this is the method behind this index and every other one on this site: the question, the tier rules and the limits.

The city sample

60 US cities across 36 states. Our list holds 142 therapy websites in them, 1 to 7 candidates each, and the screener opened and read 131. Every city had at least one candidate. The crawl below also reads sites from our list in 32 more cities, and the sites of named practices that were not on it, which is why it counts more sites. 223 of the 250 homepages it counted sit in the 60 index cities.

The AI answers

One question per city, "best therapist in [City] [State]", asked in Google AI Mode from a fresh logged out session on October 2, 2026 and October 4, 2026. We wrote down every name the answer gave, in the order it gave them. The answers also showed 315 place cards. One card, in one city, was a physical therapy or massage clinic, not talk therapy, so we kept it out of the names. Answers change, so this is a dated snapshot, not a ranking.

The homepage crawl

On October 5, 2026 we opened 275 practice homepages. 24 did not answer, blocked our crawler, or gave back a page too thin to read or not the practice we asked for. The crawler reads the homepage HTML for known scripts, links, markup and words, so a practice can use a tool without it showing here. Every share is a floor.

The relevance gate

A site is counted only when its own words or links show counseling or mental health work, unless the answer named it. The word therapy alone is not enough, because it also means physical or massage therapy. That dropped 1 of the 251 sites we read, which leaves the 250 every share on this page is out of.

Named versus unnamed

The named group is the 117 counted sites whose practice Google AI Mode listed in its own city's answer. The other 133 are everyone else. The comparison shows what the named group happened to have. It does not show why they were named, and this page never says it does.

What this page does not say

Nothing here says a practice is good, safe or right for anyone, or that any kind of therapy works. We report what the answers named and what the homepages say. We print each name exactly as the answer gave it and nothing else about that person.

How a city was coloured

By the length of the answer alone: three or more names is open, two is narrow, one or none is closed. All 60 cities named three or more, so every city in this index is open.

A comparison on one day, not a cause

Practices get named for reasons a homepage crawl cannot see: directory profiles, years of work, who links to them and what their Google profile says. Nothing on this page is a ranking factor.

221 names have no website we could match

The answers gave 340 names. We matched a website to 119 of them, 96 by testing a likely web address and the rest from our own city lists, and read 117 of those sites. Some names are one therapist, or a branch of a larger group, with no site of its own we could confirm. Those names stay in the city table, which is built from names, and are left out of the named versus unnamed comparison, which is built from websites.

78 matches were thrown out before we published

Every matched site was checked against the name the answer gave and the city it was named in. A match was removed when the site was a different business, a national homepage rather than the local practice, a domain for sale, or a page we could not read to check. We never kept a guess we could not confirm.

A few names appear in more than one city

10 of the 340 names repeat a name already given in another city. The page counts names across answers, not separate practices.

Homepage only, so every share is a floor

A fee on a rates page, an insurance list on its own page or a license on a team page is not counted. A new or home made tool reads as nothing at all.

No speed numbers, on purpose

The crawl ran from India. Response times were recorded and are left out, because from that distance they measure the length of the wire, not the speed of the site.

Small bases on some rows

The named group is 117 sites, so one site moves a named share by under a point, and a row with a handful of sites can swing by several. Quote the shares with their site counts, and never as twice as likely.