⭐ Get the Agent OS + join 3,400+ founders inside the AI Profit Boardroom → Join AIPB

GoldieBench Blog · 6 min read

AI SEO For Dentists: Six AI Models Compared Across Every Goldie Bench Page Task, Matched To Treatment Pages, Booking Flows And Schema

Landing, Web-OS and AIPB Promo scores read live on goldiebench.com, lined up per model so a practice can pick for consistency, not one peak.

Illustrated hero: a benchmark leaderboard with glowing bars beside a stylised dental practice page and a simple tooth icon, on a dark navy background with gold accents

AI SEO for dentists leans on AI models for drafting: treatment-page layouts, Dentist schema, FAQ outlines and booking-page copy.

Rather than picking a model from one headline number, this page compares how six models scored across all three Page tasks on goldiebench.com, read live on 12 Nov 2026.

Book a free SEO strategy session →

After the scores comes the work no model does for a practice, and the free option for that: book a free SEO strategy session.

Why Only The Page Tasks Matter Here

Goldie Bench runs 50 tasks across 33 models, and its models page explains the method: fixed prompts, a single attempt per model, and Julian scoring each result from 0 to 10.

Most of those tasks are games, visuals and simulations, which tell a practice very little.

The task list puts three in the Page category: AIPB Promo, Landing and Web-OS Desktop.

Those three are the nearest thing on the bench to building a dental website, so they are the only scores this page uses for model choice.

No task checks dental facts, so a high score says nothing about clinical accuracy.

The Three Page Tasks At A Glance

TaskScoredAverageWinnerPage
Landing21/268.11/10Fugu Ultragoldiebench.com/tasks/landing
Web-OS Desktop21/267.98/10Fusiongoldiebench.com/tasks/webos
AIPB Promo20/267.48/10Qwen 3.8goldiebench.com/tasks/aipbpromo

Landing asks for a modern marketing landing page in one HTML file, which is closest to a treatment page.

Web-OS Desktop asks for a working desktop with windows and apps, a test of multi-part interactive builds such as a booking flow.

AIPB Promo asks for an auto-playing motion-graphics advert, the closest match to a short practice welcome video.

Six Models Across All Three Page Tasks

ModelLandingWeb-OSAIPB PromoMean of the three
GPT-5.6 Sol8.78.48.38.47
Fusion9.09.07.58.50
Grok9.09.07.08.33
Gemini 3.6 Flash8.68.07.88.13
GLM-5.29.07.57.58.00
Claude Fable 57.28.07.57.57

Each score was read on its own task page above. The last column is our own arithmetic mean of the three, not a Goldie Bench figure.

Consistency Beats One Big Score

Fusion and Grok both hit 9.0/10 on Landing and on Web-OS, then fall to 7.5 and 7.0 on AIPB Promo.

Book a free SEO strategy session →

GPT-5.6 Sol never reaches 9.0, but it stays between 8.3 and 8.7 on all three.

For a practice that wants one model to handle every page job, that narrow range is worth more than a single peak.

It also matches the leaderboard, which tags GPT-5.6 Sol "best at Pages" and ranks it fourth overall on 8.16.

If the job is only treatment-page layouts, a Landing 9.0 scorer such as Fusion or Grok is the stronger first draft.

Which Model For Which Dental Job

Dental jobFirst choice from the scoresWhat a person still checks
Implant or whitening page layoutFusion or Grok (9.0 on Landing)A dentist approves every clinical sentence
Online booking or new-patient form pageFusion or Grok (9.0 on Web-OS)Front desk tests the flow end to end
Short practice welcome video script and scenesGPT-5.6 Sol (8.3 on AIPB Promo)No outcome promises; regulator rules checked
Dentist JSON-LD for the contact pageGPT-5.6 Sol, the steadiest Page scorerEvery value matches the page and the profile
FAQ outline from front-desk questionsGPT-5.6 Sol or Gemini 3.6 FlashA dentist signs off; no personal medical advice

For the markup row, Google's local business structured data docs ask for the most specific sub-type, and schema.org/Dentist is the one that fits a practice.

Test Two Models On One Real Page First

Benchmark scores narrow the field, but your own practice is the real test.

Run this trial before you commit to a model for the whole site.

  • Pick one treatment the practice wants to grow, such as emergency appointments.
  • Give both shortlisted models the same brief: the front-desk questions, the dentist who provides the treatment, and the booking link.
  • Ask each for a page layout and a Dentist JSON-LD block, in one attempt, the way the bench does it.
  • Have the dentist mark every clinical error, and have the practice manager mark every layout problem.
  • Keep the model with fewer clinical corrections, even if its layout needs more work, because layout is cheaper to fix.

Book a free SEO strategy session →

Repeat the trial whenever a new model tops the Page tasks, since the bench adds models often.

Where The Overall Leaderboard Fits

The leaderboard's top four are Fusion on 8.59, Claude Opus 5 on 8.27, Hermes MoA on 8.17 and GPT-5.6 Sol on 8.16.

Those averages mix every category, so a strong game builder can rank high without being the best page builder.

Use the overall rank for thinking jobs, such as grouping 200 patient questions into a sensible list of treatment pages.

Use the Page scores above for anything a patient will actually see.

What The Scores Cannot Tell A Practice

Health topics sit in the "Your Money or Your Life" group in Google's helpful content guidance, which says such content must be highly accurate and in line with expert consensus.

No Goldie Bench task measures that.

So every draft, from a 9.0 model or a 7.0 one, goes to a dentist before it goes live, and the practice checks its dental regulator's advertising rules on claims.

The Search Work No Model Replaces

  • Profiles: Google's Business Profile guidelines want one profile per public-facing dentist, not one per specialism.
  • Reviews: Google's review tips page bans offering anything in return for a review, and its Maps policy bans asking only happy patients.
  • Crawlers: OpenAI says sites that block OAI-SearchBot won't be shown in ChatGPT search answers.
  • AI answers: Google's AI features page says no special optimization is needed for AI Overviews or AI Mode beyond SEO best practice.

Get these four right and the model choice becomes a matter of saving drafting time.

When To Bring In A Team

A model can draft a treatment page in minutes, but profile clean-up, review handling, crawler checks and monthly booking reports are ongoing.

That is what Goldie Agency covers, and Julian's own published numbers on his about page include 425K+ YouTube subscribers and 1,000+ websites ranked.

Nobody can guarantee a practice a first-place ranking or a mention in an AI answer, as Google's hiring page makes clear.

Book a free SEO strategy session to decide which jobs stay with your models and which to hand over.

To learn AI SEO yourself instead, start inside the AI Profit Boardroom.

Free downloads: the 200+ AI SEO Prompts for drafting and the free link building book for local outreach.

Related reading: models that help you rank in Google AI Overviews.

Book a free SEO strategy session.

Book a free SEO strategy session →

Also On Our Network

Run your steadiest Page scorer on one treatment page this week, have a dentist review it, and book a free SEO strategy session when you want a team on AI SEO for dentists.

FAQ

Which model is most consistent on Goldie Bench Page tasks?

Of the six compared here, GPT-5.6 Sol scored 8.7 on Landing, 8.4 on Web-OS and 8.3 on AIPB Promo, the narrowest range of the group.

Does Goldie Bench test dental or medical accuracy?

No. The tasks are one-shot creative coding builds. A dentist must review any clinical content a model writes.

Which model should draft a treatment page layout?

Fusion and Grok both scored 9.0/10 on the Landing task and on Web-OS, so either is a strong first draft for layouts. A dentist still reviews the content.

Will a better model get my practice into AI Overviews?

No model can promise that. Google says AI Overviews need no special optimization beyond SEO best practice, and they only show when they add value.

The same stack Julian uses

Run this stack yourself.

Every demo on this bench was built inside the Agent Operating System — one prompt, one shot, single HTML file out. The Agent OS, the prompts, the templates, the weekly walkthroughs and 3,400+ founders shipping with it every day all live inside the AI Profit Boardroom.

3,400+founders
258documented wins
38countries
4coaching calls a week