
On October 7, OpenAI started rolling out GPT-6 in ChatGPT to its paid plans. The next day the rollout began reaching the Free and Go plans. OpenAI says more than 1.2 billion people use ChatGPT every week, so for a large share of your future patients, this is now what AI looks like.
Three parts of the release are worth understanding.
The free version got the new generation
Paid plans run GPT-6 Sol. Free and Go users get GPT-6 Luna, a smaller model from the same release. The free experience is no longer a generation behind.
Answers are no longer only text
OpenAI calls this Intelligent UI. A response can include charts, side-by-side comparisons, buttons, forms, and small tools built on the spot, like a calculator you can adjust. OpenAI says the model picks the format from the question and still answers in plain text when that works best.
It searches more carefully
OpenAI says GPT-6 is better at deciding when to look something up and more reliably finds information that supports its answer. On questions that need a web search, GPT-6 Instant starts answering 44 percent sooner on average than the previous version. These are OpenAI’s internal evaluations, not independent tests.
Why a practice should care
Think about what a prospective patient wants to know before booking. What a treatment costs. How much downtime to plan for. Whether they’re a good candidate. What results look like. How it compares with the option a friend mentioned.
More of those questions now go to an assistant first, and the answer comes back structured: a comparison, a recovery timeline, maybe a rough cost they can play with. The assistant builds it from what it can find about you, on your site and everywhere else. If that information is accurate and easy for AI to read, the answer has something good to draw on. If it isn’t, the comparison gets built from someone else’s information, and the patient walks into your consult with it.
In August I wrote about how AI assistants decide which practices to recommend. GPT-6 matters for the same reason. The answer patients get is more detailed and more visual than it was.
What I wouldn’t do
I wouldn’t rebuild your website around this, and I wouldn’t launch a chatbot because OpenAI shipped a new model. Intelligent UI isn’t a feature you need to copy. What moved is the patient’s starting point.
Start with what AI says about you now
Open a free ChatGPT account and ask what your patients ask. Cost, downtime, candidacy, and results for your three highest-value treatments, plus how each compares with its usual alternative. Then ask who the best provider for each one is in your city. Write down what comes back and which sources it cites.
Look for two things. Whether the treatment answer is generic, and whether the answer about who to see includes you, with current details.
Then get help making the right information show up
This is the part I’d hand to Local AI Registry. Assistants don’t only read your website. They pull from your Google profile, directories, review sites, best-of lists, and media across the web, and Local AI Registry works on all of those.
Start with their free AI Score. It shows which assistants name your practice, what they say, and why, scored across the same eight signals I covered in August. Claiming your record is free.
From there, their paid levels handle the ongoing work: keeping your details identical across directories, managing your Google profile and reviews, and building the pages and structured data AI reads. Their higher levels also prepare your photos and media for AI, with location and structured data attached, which is where your before-and-after galleries come in. Publish those with patient consent, then ask the Local AI Registry team how to get the right ones in front of AI when a patient asks what results look like.
Your own pages still matter. Keep treatment details, pricing approach, provider credentials, and aftercare current, and hold them to your advertising standard: no outcome guarantees, and before-and-after photos only with consent.
The limits
OpenAI says the model’s design judgment still needs work. Free users get the smaller model. An AI answer about candidacy is general information, not a clinical assessment, which is what the consult is for. And no one can promise how an assistant will answer on a given day.
Patients will sometimes get incomplete or wrong answers. That’s the best reason to make sure the information about your practice is accurate everywhere AI looks.
Look up your practice’s free AI Score at localairegistry.com.
Disclosure: I work with Local AI Registry.
Warmly,
Audrey
AI drafted, author reviewed.
Originally published on Substack.