Same dental case, four different answers: DeepSeek came out on top
In specialist healthcare, finding out which engine describes your field accurately matters more than whether “AI mentioned us”.
Key numbers
- Models evaluated: four frontier systems (GPT-4o, Gemini 2.0 Flash, Copilot and DeepSeek V3), tasked with open-ended clinical analysis of longitudinal dental cases.
- Sample: 34 standardised longitudinal periodontal cases comprising 258 question-answer pairs, scored with automated metrics and blinded ratings by licensed dentists.
- Result: DeepSeek led on faithfulness (median 0.528 versus 0.367–0.457) and on expert ratings (median 4.5/5 versus 4.0/5), without a meaningful readability penalty.
- The authors position it as an adjunct tool for medical education and research, and a candidate domain-specific agent.
Method and sample
arXiv paper “DeepSeek performs better than other Large Language Models in Dental Cases” (2025-09-02): 34 standardised longitudinal periodontal cases with 258 QA pairs, used to test four models' open-ended clinical reasoning through automated metrics plus blinded expert scoring.
What it means for brands
For the same specialist question, answer quality genuinely differs by engine. For healthcare and dental providers, knowing which engines describe your specialty accurately — and which invent things — is far more actionable than a single “AI mentions us” count.
How SourceSight handles it
SourceSight runs 10 engines and reports mentions and position per engine; specialist projects also build prompts from industry vocabulary (treatment types, materials, credentials) rather than generic phrasing alone.
Limits
The evaluation covers 34 periodontal cases and 258 QA pairs in a single specialty, so it is small-sample evidence; model versions move fast and results do not transfer to other specialties or newer releases.
References
Run it on your own project
What the same prompts return across 10 engines is only knowable by running them.
Related research
Cross-category recommendations: +36% add-to-cart on item pages, +15% on cart pages
Recommending general merchandise to grocery shoppers — a milk frother with the milk — has long been hard; LLM retrieval plus real-time cart context moved production A/B metrics.
Industries · 2026-10-07Home services: 22% of homeowners already use AI to find a provider, and 80% of businesses don't know how to be found
The local services funnel is moving upstream: before a homeowner opens a website or calls, AI has already synthesised reviews, reputation and service descriptions into one answer.