AI is everywhere. From chatbots that help with customer service to generative tools answering questions, people AI adoption continues to grow. While there’s no question that AI is transforming industries and online behavior, the technology is far from perfect. In fact, there are still plenty of challenges and limitations we need to be mindful of as AI continues to evolve. So, we’re constantly balancing a dichotomy between:

  • There is a lot of current and potential future application of AI (both for our own usage and getting our brand in front of our target audience)
  • There are still many issues with AI under the surface and the AI hype is still focused on a visionary picture of the future rather than the current reality

There’s no denying that AI adoption is increasing. Consumers are using AI tools more than ever, and businesses are eager to capitalize on marketed benefits such as increased efficiency, lower costs, and improved output. It’s a trend that’s hard to ignore, but all new technologies come with limitations, and it’s important to understand where the limits of AI lie.

AI Adoption is on the Rise but So Are the Challenges

For businesses adopting AI, it’s crucial to understand both the opportunities and the risks. People are using AI tools, but they may not be fully aware of how flawed these systems can be, especially when it comes to delivering accurate results.

The Reality Behind AI: What We’ve Learned So Far

Recent studies and findings have shed light on some serious shortcomings in AI that businesses need to keep in mind. While there are too many to go into in one article, here are some of the ones we’re regularly talking about:

  1. Generative AI Can Often be Confidently Wrong: One of the most common consumer uses of AI is Generative AI such as ChatGPT or Google’s AI Overviews powered by Gemini, but they are not as accurate and reliable as we’d like them to be. Unless there is a safety concern with your question, these platforms will almost always provide an answer, even if they don’t have the right answer. Instead of admitting they don’t have the answer, they will provide an overly confident response, leading to confusion and frustration. Just take a look at this recent research by the Columbia Journalism Review. They found the following key points, which our teams have seen in many cases with our client’s work:
    • Chatbots were generally bad at declining to answer questions they couldn’t answer accurately, offering incorrect or speculative answers instead.
    • Premium chatbots provided more confidently incorrect answers than their free counterparts.
    • Multiple chatbots seemed to bypass Robot Exclusion Protocol preferences.
    • Generative search tools fabricated links and cited syndicated and copied versions of articles.
    • Content licensing deals with news sources provided no guarantee of accurate citation in chatbot responses.
  2. They Still Hallucinate: Generative AI tools are made to provide answers based on available data, but they can sometimes create fabricated information. This is part of their “D.N.A” and what makes them create creative companions. However, if you’re looking for pure information retrieval and not a creative headline, this can cause problems. We’ve seen them provide links to pages that don’t exist, make up studies that never happened, and reference people that aren’t real.
  3. Premium Tiers of AI Tools Aren’t Always the Solution: You might think that paying for a premium / paid subscription guarantees better accuracy, but that’s not always the case. 
  4. AI Bypasses Important Protocols: Many AI tools don’t always respect website owners’ preferences, particularly when it comes to the Robot Exclusion Protocol (REP). REP is a set of rules that allows website owners to provide clear directions on what should not be viewed by outside crawlers. However, most AI tools have ignored this. There are a lot of proposals being generated for tools to control how these AI platforms access and use content from websites, but none are fully supported, and the platforms have already captured the entire web in their training data.
  5. Licensing Doesn’t Guarantee Accuracy: Just because an AI tool has access to licensed content doesn’t mean it’ll use it correctly. Even when AI platforms have deals with news outlets and data providers, there’s no guarantee that the content they generate will be accurate or properly cited. This can lead to misinformation slipping through the cracks.
  6. Privacy Concerns: For the last few years, we have been heavily focused on changes in the healthcare landscape tied to privacy concerns around patient data. Federal regulations were focused on people simply visiting a healthcare website and viewing a page about a condition or doctor, even without inputting specific information on their condition. What does the privacy landscape look like in relation to the extreme level of detail that people go into with their questions for Generative AI platforms? Will there be issues? Will regulation catch up? There is a lot to figure out here…
  7. Copyright Concerns: If there is a lot to figure out with privacy concerns, copyright is an even bigger question. There has been some progress on this topic from the U.S. Copyright Office, but significantly more clarity is needed. How long will this take and what will the implications be? 

Moving Forward with AI Thoughtfully

While there are many issues and challenges, it shouldn’t stop you from testing, learning, and iterating with AI. As adoption continues to grow, it’s important to stay up-to-date with its evolving capabilities.

The future with AI is bright, but it’s important to stay aware of its growing pains. As we continue to experiment with AI, we will continue to test, learn, and adapt so we can use it in ways that benefit both businesses and consumers alike.