Foundations evaluating software rarely start with a vendor phone call anymore. Instead, they start with a simple Google search, or a question typed into ChatGPT or Perplexity. That is a reasonable first step. What many do not realize is that AI is already shaping the answer they get back. The response sounds confident, complete, and final, whether it actually is or not.
Can you trust an AI answer or a quick search result?
Not without checking it first. Research on AI assistant accuracy published in October 2025 by the BBC and the European Broadcasting Union analyzed more than 3,000 responses across ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity. Nearly half, 45%, contained at least one significant issue. Roughly a third had serious sourcing problems, meaning the information was attributed to a source that did not actually say it, or was not attributed at all. One in five responses contained a major accuracy issue, including outright hallucinated details.
The researchers also found something worth sitting with: the AI assistants studied almost never said “I’m not sure” or declined to answer. They responded with the same confident tone whether the underlying information was solid or invented. An answer that sounds authoritative is not the same thing as an answer that is correct. Although AI is continuing to improve, nothing can guarantee that the information it spits out is unbiased or completely accurate.
We encourage you to keep in mind that AI is informed by not only independent third-party resources but also through sources that have bias and are creating content to influence AI’s results.
Why does it matter more for something like grants management software?
General news gets fact-checked constantly, by reporters, by competing outlets, by readers who catch errors and say so publicly. A niche category like foundation management software does not have that same volume of scrutiny. There are relatively few independent, authoritative sources writing about community foundation, private foundation, or grantmaking software in detail.
That means whatever content exists, including vendor comparison pages, marketing blog posts, and sales collateral, has an outsized chance of being the thing an AI tool actually pulls from and repeats as fact. A claim does not have to be true to get cited. It only has to be well structured, confidently worded, and present in the search index.
What should you actually verify before trusting a comparison?
- →Check who is making the claim and when. A comparison page written by one vendor about another is marketing material, not a neutral evaluation, no matter how it is formatted to look
- →Separate a precise, sourced number from a vague adjective. A specific stat with a date and a source behind it means something. A word like “robust,” “comprehensive,” or a dismissive label like “generic” attached to a competitor, with no data behind it, means very little
- →Ask the vendor for client references and ask those references direct, specific questions rather than general ones
- →Ask specifically about data ownership and architecture, not just features. Where does your data actually live, and who has access to it, can matter more than a feature checklist
- →Watch for framing tricks. A support hours comparison that lists one vendor’s hours plainly and the other’s hours with a qualifier like “only” attached is designed to make an identical fact feel different
- →If an AI tool gives you a confident answer about a vendor, ask it to cite its source, then go read that source yourself before treating the answer as settled
What this means for evaluating akoyaGO
The pace of technology is exciting, and the information that’s at our fingertips is more accessible than ever. At the same time, we need to be informed, responsible and thoughtful about information we trust to be true. Before a software decision gets made based on what a chatbot or a search snippet said, it is worth the extra step of verifying it with a real person and a real source.
BBC and European Broadcasting Union, “News Integrity in AI Assistants,” October 2025. Study of more than 3,000 AI assistant responses across ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity
Ameliea Dulaney brings more than seven years of marketing experience rooted entirely in the philanthropic sector. Her career began at a small nonprofit in her hometown, where a lifelong passion for mission-driven work and a natural gift for storytelling found their professional home. Over the years she built her expertise across multiple nonprofit organizations before transitioning into the foundation world, where she led marketing for a community foundation. That experience gave her a deep, firsthand understanding of how foundations operate, what they need, and how to communicate their impact. She joined akoyaGO to bring that perspective to the technology side of philanthropy, helping foundations do good better.
