Published Research
Physical Therapy AI Citation Study
What makes AI recommend one physical therapy practice over another? To find out, we analyzed over 7,000 AI responses across 21 U.S. cities and 3 major AI platforms. This is what we found.
Key Takeaways
- ✓75.7% of AI citations go to independent practices, not large chains or hospital systems.
- ✓Domain authority has essentially zero correlation with AI citations (r = 0.002). Backlinks don’t matter here.
- ✓Website structure is the strongest predictor of AI recommendation — the same finding from our chiropractic study.
- ✓Practices with dedicated condition pages (knee pain, post-surgical rehab) get cited far more than those with generic service pages.
- ✓Only about 30% of practices maintain visibility across consecutive queries. Consistency is rare — and valuable.
On this page
Study at a glance.
AI responses analyzed
U.S. cities studied
AI platforms tested
Citations to independent practices
Domain authority correlation (r)
Heading structure advantage
Study overview.
We designed this study to answer a practical question: if someone asks an AI tool for the best physical therapist near them, what decides which practice gets named?
We tested across 21 U.S. cities of different sizes. In each city, we asked ChatGPT, Perplexity, and Gemini the same sets of questions about physical therapy. We varied the wording, the conditions mentioned, and the type of query. Over 7,000 total responses were collected and analyzed.
For every practice that appeared in an AI response, we recorded what it looked like online. We measured website structure, content depth, domain authority, review counts, specialty pages, schema markup, heading hierarchy, and more. Then we looked for patterns.
This study follows the same pre-registered methodology we used in our chiropractic AI citation study. That means we decided what we were measuring before we collected data — not after. This prevents cherry-picking and keeps the results honest.
Key Finding
Independent practices dominate AI citations.
This was the most striking finding. 75.7% of all AI citations went to independent physical therapy practices. Not hospital-owned clinics. Not national chains. Independent practices.
This goes against what most people expect. Big chains have more locations, more brand recognition, and bigger marketing budgets. None of that mattered. AI models consistently preferred practices with deep, specific information about their services.
Why? Independent practices tend to have websites that describe what they actually do. They list the conditions they treat. They talk about their approach. They have pages about specific problems like knee pain after surgery or sports injuries in runners. Chain practices tend to have cookie-cutter websites that say the same thing across hundreds of locations.
AI reads content. When one website explains exactly how it treats rotator cuff tears and another just says “we offer physical therapy services,” AI knows which one to recommend.
Of AI citations go to independent practices
Of AI citations go to chains and hospital systems
Key Finding
Website structure is the top predictor.
For the second time in our research, website structure came out as the single strongest predictor of AI citations. How your website is organized matters more than any other factor we measured.
What do we mean by structure? Clear navigation. Logical page hierarchy. Sequential heading tags (H1, then H2, then H3 — in order). Pages that each focus on one topic. Information that is easy for both humans and AI to scan and understand.
Practices with proper sequential heading hierarchies showed a 2.8x advantage in AI citations. That is the same multiplier we found in the chiropractic study. This pattern is consistent and strong.
Think of it this way: AI reads your website the way a new patient would skim it. If the information is scattered, buried in paragraphs, or hidden behind confusing navigation, AI skips you. If it is clean and organized, AI uses you as a source.
Citation advantage from proper heading structure
Consistent across both studies
Key Finding
Dedicated condition pages are critical.
Practices that got recommended by AI had one thing in common: dedicated pages for specific conditions. Not a single page listing every service. Individual pages for knee pain, shoulder injuries, post-surgical rehab, sports injuries, back pain, balance disorders, and so on.
This makes sense when you think about how people ask AI questions. They don’t ask “tell me about physical therapy.” They ask “who is the best physical therapist for knee pain after ACL surgery near me?” AI needs to find a page that matches that specific question.
A practice with a detailed page about ACL rehabilitation — what the process looks like, how long recovery takes, what techniques are used — is far more likely to be cited than a practice with a generic “services” page that briefly mentions knee injuries.
The pattern is clear: specificity wins. The more specific your content is about the conditions you treat and the patients you serve, the more AI trusts you as the expert to recommend.
Key Finding
Each AI platform picks different winners.
ChatGPT, Perplexity, and Gemini do not agree on who to recommend. Each platform uses different data sources, different weighting systems, and different criteria. A practice that ranks well on one platform may be invisible on another.
This mirrors what we found in the chiropractic study. Cross-platform overlap is consistently low. If you optimize only for ChatGPT, you are missing the patients who use Perplexity or Gemini.
Only about 30% of practices that appear in one AI response maintain visibility in the next query — even on the same platform. AI recommendations are less stable than Google rankings. This means that broad, consistent optimization matters more than chasing any single platform.
The takeaway: you need to be visible everywhere. A strategy built for one AI model is a strategy that misses most of the opportunity.
Key Finding
Third-party mentions matter more than you think.
AI doesn’t just look at your website. It pulls information from everywhere — Reddit threads, YouTube videos, local directories, review sites, and community forums. A significant portion of AI answers included information from these third-party sources.
This means that your online presence beyond your own website matters. If patients talk about you on Reddit, if you have YouTube videos explaining exercises, if local community sites mention you — AI notices. These mentions serve as independent validation of your expertise.
This does not mean you should spam forums or buy fake reviews. AI is good at detecting low-quality signals. What it rewards is genuine, organic mentions — real people talking about real experiences with your practice.
The practical step is simple: be worth talking about. Do good work, create helpful content, engage with your community. The mentions will follow, and AI will find them.
Patterns that hold across both studies.
Our chiropractic and physical therapy studies were run independently, in different cities, at different times. But the same patterns kept appearing. This makes us confident these are real signals, not noise.
Website structure is always #1
In both studies, how a website is organized was the strongest predictor of AI citations. This held across all three AI platforms, all cities, and all query types.
Domain authority is always irrelevant
Chiropractic showed r = -0.05. Physical therapy showed r = 0.002. Both are effectively zero. Traditional authority signals do not translate to AI search — in either industry.
Heading hierarchy gives a 2.8x advantage
The exact same 2.8x multiplier appeared in both studies. Practices with sequential, well-organized heading structures are cited nearly three times more often.
Cross-platform overlap is low
Each AI platform picks different winners. This finding was consistent across both industries. Optimizing for a single platform is not enough.
Specialty pages are essential
Whether it's condition pages for chiropractic or condition pages for PT, dedicated service-specific content is a prerequisite for AI visibility.
Almost no one is ready
Less than 2% of practices in either industry have the right setup. The opportunity for early movers is wide open — and it compounds every month.
What this means for your PT practice.
If you run a physical therapy practice, here is the short version: the way patients find healthcare is changing. More people are asking AI for recommendations instead of searching Google. When they do, AI picks 3 to 5 names. If you are not one of them, that patient goes somewhere else.
The good news is that what AI looks for is straightforward. It is not about spending more money. It is not about getting more backlinks. It is about organizing your website clearly, creating pages for the specific conditions you treat, and making sure AI can read and understand your information.
The better news is that almost nobody in physical therapy is doing this yet. The first practice in each market to get this right will have a lasting advantage. AI models learn over time. Once they start recommending you, that recommendation gets reinforced with every positive signal.
We built our entire service around these findings. Every recommendation we make comes directly from this data — not from guessing, not from what worked on Google, but from what we measured actually works in AI search.
Frequently asked questions.
How was this study conducted?
We queried ChatGPT, Perplexity, and Gemini across 21 U.S. cities using varied question formats about physical therapy. Over 7,000 responses were collected. For each cited practice, we measured website structure, content depth, domain authority, reviews, schema markup, heading hierarchy, and more. The methodology was pre-registered before data collection began.
Why does domain authority not matter for AI search?
Domain authority measures how many other websites link to yours. It's a proxy for trust in traditional search. AI models process content differently — they evaluate the quality, specificity, and structure of the information itself rather than relying on external link signals. A small practice with excellent, well-organized content outperforms a large chain with a high-authority but generic website.
Is this study peer-reviewed?
This is proprietary industry research, not an academic journal submission. However, we follow rigorous standards: pre-registered methodology, large sample sizes, multi-city scope, and transparent reporting of our findings — including null results like the domain authority correlation. We publish the data because we believe the industry needs honest, evidence-based guidance.
How do these findings compare to the chiropractic study?
The core findings are remarkably consistent. Website structure is the top predictor in both. Domain authority is irrelevant in both. The 2.8x heading hierarchy advantage appeared in both. Cross-platform divergence is present in both. The main PT-specific finding is the 75.7% independent practice citation rate, which reflects how AI values specialization and content depth over brand scale.
Can I use this research to optimize my own website?
Yes. The findings point to clear action steps: organize your website with a clean structure, create dedicated pages for each condition you treat, use sequential heading hierarchies, and ensure your content is specific and detailed. That said, execution matters — and the nuances of how to do this effectively across all three AI platforms is where professional implementation makes the difference.
Want these findings applied to your practice?
Every recommendation we make is backed by this data. Start with a free strategy call to see how AI currently handles your market.