The AI Invisibility Index

Why we ran this study

45% of US consumers now use AI tools to find local services, up from 6% a year earlier, and mental health is the most sensitive local search anyone makes. AI assistants answer those questions by reading practice websites. We wanted one number nobody had: how much of the profession can the machines actually see?

So we pointed our website readiness engine, the same one behind our free scan, at an independent, nationwide sample and let the data speak.

What we found

Most practices are invisible. 77.9% of the 962 practices we could grade fall below "AI-readable" on the rubric: their sites load, but what an AI can extract from them is too sparse, too thin, or too broken to quote confidently in an answer.

The commonest failure is the simplest one: not enough words. 87.4% of practices have key pages under the content depth an assistant needs. A services page listing six specialties in six bullet points gives an AI nothing to say about any of them.

The new conventions have not arrived. 65.7% have no llms.txt. 15.6% carry no structured data describing the practice at all.

Speed is a profession-wide problem. The median mobile PageSpeed score is 57. Only 5.2% of measured sites reach Google's "good" band. And 18 practices, about 2%, actively block AI crawlers in their robots.txt, telling the tools their future clients use to go away.

No state passes. The best state in the country, Colorado, still has 55% of its practices below the line. Alabama and Kansas sit at 95%.

Every state, ranked

Each row is one state, worst first. The middle number is how many practices we scanned there, and the bar shows what share of them fall below AI-readable. So the top row reads: of the 20 Alabama practices in the sample, 95%, which is 19 of the 20, are invisible to AI search. A shorter bar is better, and no state's bar is short.

Methodology

Sample: 984 independent therapy practice websites, up to 20 per state across all 50 states, drawn from public Google Maps listings for therapist and counseling searches in each state's three largest cities and randomised with a fixed seed. National chains, directories, telehealth platforms, and every practice with any relationship to Moonraker were excluded. Scans ran 14 to 15 September 2026 with Moonraker's production readiness engine: reachability, HTTPS, metadata, structured data, robots.txt and AI-crawler policy, llms.txt, sitemap and per-page content depth, Google Business Profile linkage, and Google PageSpeed. 982 of 984 sites returned usable results; 962 could be fully graded on the readiness scale.

Limitations, honestly: the sample skews toward practices visible in Maps in larger cities, which most plausibly overstates the profession's readiness rather than understating it. Readability grading uses our published rubric, in which "AI-readable" requires machine-extractable practice information across the site's core pages. Speed scores are lab measurements and move a few points between runs.

The dataset

State-level aggregates are public under CC BY 4.0: download the CSV. Cite as "Moonraker AI Invisibility Index, September 2026", linking this page. Journalists: for state-level breakdowns, additional cuts, or comment, contact scott@moonraker.ai. We do not publish per-practice results.

If you run a practice

The gap is the opportunity: with four in five practices invisible, the fixes are unusually decisive. Run the same checks we used on your own site in about a minute with the free AI readiness scan, or check whether the web agrees about your practice details with the listings check.