How to Choose a Chatbot in Mid-2026
I get one question a lot in my consulting: which AI chatbot should I use? There is no “right” answer, but the best practices in making your choice have actually changed quite a bit in the past year.
An “AI LLM chatbot” is broadly a chat interface to an artificial intelligence system trained on a large language model. In plain English, it’s the website (or desktop software) where you ask questions back and forth and get human-sounding answers. Everyone knows ChatGPT, but there are lots of other choices.
The biggest change since mid-2025 has been that pretty much all of the chatbots now work similarly, and most of them work pretty well. Any choice will be OK. If you read no further, choose the cheapest paid option.
Key considerations for nonprofit AI chatbot choice include (in my suggested order of importance): cost, integration with the rest of your technology, security, effectiveness for your particular types of tasks, ease of team information sharing, ethical standards, and level of functionality required. I’ll address each factor separately except the two that are the same across all the major AI chatbots: security (high for the U.S.-based ones) and level of functionality (uniformly high among the AI leaders).
Cost
You would think more organizations would pay attention to cost, but the tech is so new that I find a lot of community benefit organizations ignore cost. One super important note: the top AI chatbots have deep nonprofit discounts! Google Gemini is free with Google Workspace for certain licenses (but not all functionality… more on that later). Claude Team and Enterprise are 75% off for nonprofits, and ChatGPT is just $8 per user per month for nonprofits. Even Microsoft offers a modest 15% discount for nonprofits. You would be crazy to miss this savings.
Keep in mind that you may not need AI chatbot access for every single employee either. Start with the needs, pilot a small group, and roll out only as necessary. I do think AI will become more important as time progresses, but you don’t need to break the bank right now to get to that point. Even $8 per month means $96 per employee, although I would argue that a $50,000 employee with 20% benefits cost means you don’t have to get much productivity increase to see a return on investment — only a 0.16% increase, or a savings of 3.2 hours per year, provides a positive return.
Integration
Most AI tools now integrate with email, calendars, documents, spreadsheets and more. The advantages of these integrations often outweigh other benefits, so I often recommend sticking with the AI tool already built by your primary provider. That means using Google Gemini if you’re a Google Workspace customer and Microsoft Copilot if you’re a Microsoft shop. One big caveat: Microsoft Copilot is both less effective and much more expensive ($25 per user per month versus $8 for others), so Microsoft Office users may want to consider an alternative. Anthropic’s Claude AI generally seems to work well with Microsoft products.
The integrations work in a variety of important ways. In spreadsheets, you can use AI to create complicated formulas or to analyze data. In Google Slides or PowerPoint, you can create entire presentation drafts from outlines or notes. Even in Gmail and Calendar, the desktop versions of some of these tools allow you to search by concept and even book meetings. Enabling the AI to click for you makes it easier to do unfamiliar tasks instead of having to puzzle through the often cryptic help section.
I suspect that the integrations will get better and more complicated. This path suggests a future-proofing benefit of learning Gemini or Copilot because I predict that their internal integrations will work better over the long term than third-party AI tools can accomplish.
Effectiveness for Your Tasks
At this point, the top AI chatbots all do everything fairly well. That said, certain AI chatbots lend themselves better to certain tasks:
Microsoft Copilot’s “agentic” AI work (having Copilot take actions for you) tends to work in a more coordinated way within Microsoft products.
Claude tends to write cleaner, more human-sounding prose and provides better summaries.
I have gotten somewhat better results using Gemini within Google Sheets and Copilot within Microsoft Excel than through using another tool, although you can still use whatever tool you have for analysis.
Claude and Gemini write computer code with better comments and instructions if you’re a beginner.
Claude has some useful extras such as the ability to type a slash (“/”) and get both a list of built-in commands and access to your own Claude Artifacts.
There are certain tasks that are just plain easier within a highly specialized tool as well. For example, if you want to remove the background from a photo, the AI tools within Adobe InDesign and Canva frankly work better and faster than the standard AI chatbots do.
My best recommendation is to start with the work you need to do and then to experiment with your existing tool to see if you can make it work rather than buying an additional tool. Don’t hesitate to lean on the chatbots themselves to explain which features of which chatbots best apply to your particular problem!
Information Sharing
All of the major chatbots (Gemini, Copilot, Claude and Perplexity) now have team plans. With the exception of Perplexity, they also meet all of the most important team requirements: access to the latest AI models for better performance, ability to access “agentic” models that take action such as booking meetings on your behalf, and the ability to create and manage team workspaces, SSO (single sign on), audit logs, and billing. They also meet security concerns: none of the team plans allow the AI models to train on your business data, and they provide overall team security, license management and admin controls.
Perplexity offers the most basic team tools and collaborative spaces. While Gemini, Copilot, and Claude all let you create "workspaces" with shared files, the ways teams interact with these workspaces are quite different:
Claude Cowork & Artifacts: Claude’s interface is designed for active co-creation. “Artifacts” generate dedicated, side-by-side windows for documents or code, making editing a breeze. You can share these across your team using the “Share” button. You can also share prompts and more within Claude Projects. The Claude Cowork tool allows you to schedule multi-step, remote agent tasks that run in the background and have results waiting for you later, and you can share these within Claude Projects as well.
Copilot Project Hubs, Notebooks & Scout: Microsoft’s Copilot Notebooks shift the focus from a sidebar chat to persistent project folders. You bundle your source files, notes, and emails into a Notebook, and Copilot actively operates within that shared container. Copilot’s “Prompt Gallery” has a dedicated “Team Prompts” tab that allows you to share prompts with everyone or only specific people or groups. Combined with Microsoft Scout, Copilot acts as a proactive personal agent tracking tasks across your apps. IMPORTANT NOTE: Scout costs extra! Teams pay more for Scout by usage. Copilot Cowork also requires additional spending for “credits” to do these taks.
Gemini Projects and Gems: Gemini’s core workspace strength is its ability to handle huge files (with a large so-called “token window”) combined with native multimodality. You create a “Drive Project” that stores everything you need. Your team can drop hours of video or hundreds of PDF pages directly into a shared Gemini space and query with high retrieval accuracy. Gemini stores custom profiles in “Gems” that can be shared across your team. Sharing prompts is a bit more complicated with Google Vertex.
A few final, notable advantages and disadvantages of Claude:
Claude recently introduced a self-service HIPAA compliance setup for nonprofits who need this security for sensitive data, but Gemini and Copilot require HIPAA provision at the workspace level.
Claude generally produces the most natural, human-sounding writing.
Claude defaults to “view” mode only when you share an Artifact. The owner must manually adjust the collaborator's permission state to Editor to change it, which can be annoying.
Ethical Standards
If you see me avoiding a discussion of Grok and DeepSeek, two of the highly effective and lesser known chatbots, there are two reasons: 1) they are more oriented to computer programmers, and 2) I have ethics concerns about them.
We’re in the Wild West period of AI development right now, and people with ethical concerns about AI have good reason to be skeptical. Among a host of problems with AI, the technology uses enormous amounts of energy, arguably makes us dumber, contains underlying bias against historically marginalized groups, and poses major security risks. I don’t think nonprofits can avoid using AI anymore, but you can consider ethics when purchasing a system.
There are some subtle but important differences in AI ethics among the major competitors:
Microsoft’s Copilot seems very focused on preventing “hallucinations,” or the problem of AI making stuff up. It tends towards the conservative but also sometimes seems less effective as a result. Other areas of ethics don’t appear to be a focus for development, although I haven’t seen major problems in comparison to Gemini or Claude.
Google’s Gemini appears most focused on security and only somewhat focused on ethics.
Anthropic’s Claude claims to be the “ethical AI,” although I see some ethical problems with its output. Nevertheless, at least it is a theoretical focus for the company.
OpenAI seems to care about ethics only to the extent that they may get in trouble for ethical problems. Of the top AI chatbots, I believe ChatGPT is the least ethical.
You should think about getting an AI policy ASAP, and it’s worth some discussions about what level of ethics and what specific ethical concerns you want your nonprofit to address.