Best AI Tools for Fundraising: Part 1 – Comparing ChatGPT, DeepSeek, and Gemini for Donor Communications

Generative AI is continuing to grow in terms of its applications and products. This is particularly the case in Advancement with teams ranging from advancement services to annual giving to major gift officers using generative AI to work faster, spur creativity, and get more done. With all the options, which AI is the best to use? This will be a three-part series focusing on content in week one, followed by data analysis and images.

I tested three generative AIs—ChatGPT, DeepSeek, and Gemini—to determine the best AI tools for fundraising and which delivers the best results for Advancement teams creating content. The “test” was using the exact same prompt to determine which AI produced the results that met the prompt, provided accurate information, and usable output. As a secondary measure, I was also focusing on the ease of using the AI in terms of time and supplemental information. The following is a breakdown of the AIs, prompts, and the results.

The Rivals

OpenAI’s ChatGPT is the breakthrough AI, which gained mass appeal and paid users quickly. As of January 2025, there are 300 million active users per week. Gemini has been included in Google users’ daily interactions with the suite of products, growing in popularity over the last six months, providing summaries at the top of Google searches and being added to Gmail. DeepSeek is the newest AI and making news with the comparatively low setup cost. It has also experienced large-scale malicious attacks on its services and exposed user data. All three sites have logins and a similar user interface. Gemini has the advantage of being a part of Google and embedded throughout the Google suite, including browsers.

While these three platforms are the focus of the comparison, there are other generative AI options available like Microsoft’s Copilot, which is preferred by many individuals and institutions due to the enhanced data protection.

The Test

To make the comparison as accurate as possible, I use the same approach with each AI, including the following:

Basic, unpaid access
Desktop version
Same prompt, with no additional information provided

I focused on creating a communication to donors, specifically a direct mail solicitation to young alumni about scholarships.

The Comparison

The communications prompt used across all three AIs was:

Write a direct mail solicitation from a university to alumni from the last 10 years. The signer is a fellow alum. The focus is on scholarships, showcasing the impact at the university.

As I didn’t specify a university, fact-checking was not necessarily needed. The content that ChatGPT and Gemini provided was a good first draft with “[University Name]” or “[Number] of students” to add in the unique characteristics of the institution. Given the scholarship focus, there were similar open fields to add a student’s name and add a brief story to personalize the content. There were also P.S.’s and space for giving links.

The content generated from DeepSeek was different. Using the exact same prompt, it took three reentries of the same information to produce results. When the results did come through, many of the aforementioned open fields were included, along with the space for links and a P.S. The points that differentiated DeepSeek include the following:

A summary of assumptions that the solicitation content is based on that could be used to explain why the content notes specific things like “since the signer is a fellow alum, using “we” and “our university” would create a sense of community”.
In the content itself, instead of providing an open field for a students’ name and story, a fictitious story of Jessica was included as well as fictitious stats.
An ask ladder was included and a recurring giving option was referenced along with more persuasive content.
A “why this works” summary.

Overall, while the DeepSeek content will require additional editing time and more detailed prompts to remove the fictitious information compared to the output from ChatGPT and Gemini, the structure and content was more engaging. If speed is the goal, with limited edits, both ChatGPT or Gemini are suitable options.

Data is different. It takes longer to review, and small amounts of information introduced into a dataset, like the fictitious Jessica, can have a significant impact on the results. More information on the data analysis results will be provided in Part 2!

Interested in participating in an upcoming SFG blog? Let us know!

Shaff Fundraising Group is a consulting firm specializing in fundraising, marketing, and analytics. We take pride in our independent approach, free from technology affiliations with SaaS and other companies. This allows us to provide objective, solutions-oriented support to our client partners and the broader fundraising and engagement community.