Introduction
The AI market has matured significantly. In the early days (2023-2024), "ChatGPT" was synonymous with "AI." Today, in 2026, we have a diverse ecosystem of specialized models, each with distinct strengths.
For a busy physician, testing every new tool is impossible. This guide breaks down the major platforms, comparing their features, pricing, and best use cases in a medical context. For more background on these technologies, see our guide on getting started with LLMs.
The Big Three: General Purpose LLMs
These are your daily drivers — the chatbots you talk to for writing, reasoning, and planning.
1. ChatGPT (OpenAI)
- Current Model: GPT-5 (Hypothetical/Current Leading Model) / o1-series (Reasoning)
- Best Feature: Reasoning Capability. The "o" series models are designed to "think" before they speak, making them superior for complex logic puzzles, differential diagnosis, and coding.
- Pricing: Free tier available; "Plus" subscription (~$20/mo) recommended for access to advanced models and file uploads.
- Clinical Use Case:
- Complex Logic: "Here is a complex patient scenario with conflicting guidelines. Walk me through the decision-making process for Treatment A vs. Treatment B."
- Data Analysis: Uploading a spreadsheet (de-identified!) to create instant charts and p-values.
2. Claude (Anthropic)
- Current Model: Claude 3.5 Opus / Sonnet
- Best Feature: Writing Style & Context Window. Claude is widely regarded as the best writer. It produces natural, nuanced prose that sounds less "robotic" than GPT. It also has a massive context window, meaning you can paste entire textbooks or long PDF guidelines into it.
- Pricing: Free tier; "Pro" subscription (~$20/mo).
- Clinical Use Case:
- Grant Writing: Drafting specific aims pages or letters of intent.
- Summarization: Uploading a 50-page consensus guideline and asking, "Extract the treatment algorithm for Stage III disease."
3. Gemini (Google)
- Current Model: Gemini 1.5 Pro / Ultra
- Best Feature: Ecosystem Integration. Gemini lives inside your Google Workspace. It can read your emails, check your Drive files, and put answers directly into a Google Doc.
- Pricing: Bundled with Google One AI Premium.
- Clinical Use Case:
- Admin Efficiency: "Find the email from Dr. Smith about the tumor board schedule and add those dates to my calendar."
- Multimodal Analysis: It is excellent at analyzing images and video (though not for diagnostic radiology use).
The Research Assistants: Search & Synthesis
These tools are different. Instead of generating text from training data, they search the live web and synthesize answers with citations.
4. Perplexity AI
- What it is: A "Answer Engine." It replaces Google Search.
- Best Feature: Citations. Every sentence typically has a footnote linking to the source.
- Clinical Use Case:
- Quick Drug Queries: "What is the renal dosing adjustment for Vancomycin?" (It will link to UpToDate or similar sources).
- Literature Scans: "What are the latest papers on immunotherapy for glioblastoma published in 2025?"
5. NotebookLM (Google)
- What it is: A "Personal Knowledge Base." You upload your documents (PDFs, papers, notes), and it answers questions only based on those documents.
- Best Feature: Grounded Truth. It virtually eliminates hallucinations because it is restricted to the sources you provide. It also creates "Audio Overviews" — podcast-style discussions of your papers.
- Clinical Use Case:
- Board Study: Upload 20 key papers and ask it to quiz you.
- Guideline Adherence: Upload your hospital's specific protocols and query them: "What is our protocol for neutropenic fever?"
6. Elicit
- What it is: An AI research assistant for literature review.
- Best Feature: Matrix View. You ask a question, and it finds relevant papers, extracting key data (N, methodology, findings) into a structured table.
- Clinical Use Case:
- Systematic Reviews: Rapidly screening hundreds of abstracts for inclusion criteria.
Coding & Development Assistants
For the physician-scientist or the "med-tech" builder.
7. GitHub Copilot / Cursor
- What it is: AI that writes software code.
- Best Feature: Autocomplete for programming.
- Clinical Use Case:
- Data Science: Writing Python/R scripts to analyze your research database.
- App Building: Creating simple web apps for patient surveys or internal department calculators.
Comparison Table
- ChatGPT - Strength: Logic & Reasoning. Weakness: Can sound robotic. Best for complex problem solving.
- Claude - Strength: Writing & nuance. Weakness: Web search is weaker. Best for writing papers & letters.
- Perplexity - Strength: Live info & sources. Weakness: Not for creative writing. Best for quick answers & verification.
- NotebookLM - Strength: Grounding (no hallucination). Weakness: Limited to uploaded files. Best for study & specific protocols.
- Gemini - Strength: Google integration. Weakness: Can be overly cautious. Best for Google Workspace users.
Conclusion: What Should You Pay For?
If you only pay for one subscription:
- Choose Claude Pro if you do a lot of writing, grant proposals, or heavy reading/summarizing.
- Choose ChatGPT Plus if you need advanced data analysis, image generation, or deep logical reasoning.
- Choose Perplexity Pro if you mostly need to look up facts and find sources quickly.
Most physicians find that a combination of a "Writer" (Claude) and a "Searcher" (Perplexity) covers 95% of their needs.