My experience getting daily news from AI and why it hasn't worked so far

While AI models offer many advantages in various fields, they also pose challenges and ethical considerations, especially in daily news gathering.

Using AI to get daily news may seem like an innovative step towards getting information quickly and accurately. Despite the great developments in this field, I have noticed that the experience of relying on AI to get news updates has not achieved the expected success so far. There are challenges related to accuracy, updating information, and understanding the context of the news. AI chatbots such as ChatGPT, Chatsonic, Perplexity, Copilot, and Gemini reveal inconsistencies when bringing daily news updates. Check out Beyond ChatGPT: A Vision for the Future of Chatbots and Generative AI?

Can You Use Generative AI for Your Daily News? Comparing the Best Models

I tested the best generative AI chatbots for my daily news fix, and I got a variety of responses. AI chatbots like ChatGPT, Chatsonic, Perplexity, Copilot, and Gemini all provide varying responses when delivering news updates due to differences in interpretation, constraints, and contextual background.

Here's how they stack up.

ChatGPT (Free)

The free version features: ChatGPT Now with GPT-4o, OpenAI’s updated AI model. While the free version of GPT-4o can connect to the internet, it has a limit of 16 questions in a 24-hour period, after which it reverts to the older GPT-3.5 model.

Normally, checking daily news with GPT-4o is useful. Since you’re only asking one question, like “Please give me today’s headlines” or something similar, you won’t be using much of the free customization of the stronger model. Of course, you can search for specific information, but that will impact your daily limits. It’s also worth noting that if you’ve used your daily quota of GPT-4o messages, asking GPT-3.5 for a daily news update is pointless, since it doesn’t have an active internet connection, and its training data cutoff is April 2023.

Asking for daily news or some of the day’s top headlines was mostly helpful to ChatGPT, though there were some obvious limitations when it came to selecting sources, detailing those sources, and providing a range of different sources. For example, ChatGPT would quickly provide some of the leading headlines, which is great, but the URL provided as a source never linked to any specific page, even if it referred to a specific event.

It also occasionally suggests old news, which is not a limitation of the data cut point. A prime example can be found in the image above, where ChatGPT suggests that “cybersecurity in the age of AI remains a critical topic, with industry experts like Amazon’s CSO Steve Schmidt discussing the challenges and strategies for protecting data,” citing GeekWire As a source. A closer examination revealed that GeekWire published this story — on June 15, 2024, about a month ago.

Chatsonic

Like ChatGPT, it responded Chatsonic Decent for headline requests and daily news updates. It also covered a good range of topics, such as current events, politics, business, technology, and entertainment. I also liked that it provided a short summary of the news story, which is helpful for knowing what you want to read.

However, the scope of Chatsonic’s sources was a bit questionable. For example, one search used the Hindustan Times (a generally centrist daily newspaper based in Delhi, India). While that newspaper is on the other side of the world, the published source is outdated. The second source was the Dryden Wire, “Northwest Wisconsin’s #1 digital news outlet.” Needless to say, that’s a bit more niche.

Now, I can see why this might have happened. On this particular day, one of the headlines was about the Wisconsin primary, which is scheduled for August 13, 2024. Chatsonic posted a headline from that event, despite it being relatively small in terms of its impact on the US presidential election (scheduled for November 2024). Check out the best options for OpenAI’s ChatGPT for mobile.

Perplexity

The daily news provided by the Perplexity chatbot was also somewhat helpful; every time I asked about the daily news, it provided a bunch of headlines from the past few weeks, some of which were literally breaking news.

For example, a specific question, “Can you give me today’s headlines from the US?” gave a headline referring to the shooting at a Donald Trump campaign event that occurred on July 13, 2024 — a full month before my search. Conversely, the next headline addressed the United Auto Workers union filing lawsuits against Donald Trump and Elon Musk, which occurred hours before my request (on August 13, 2024).

A search for UK headlines yielded similar and mixed results. Headlines included a tragic attack and the sentencing of a controversial figure, both of which occurred weeks earlier.

With Perplexity, it seems like you might have to try a few different prompts to get the most accurate news reports. When you found the right combination, despite the wrong headline story, the rest of the headlines were accurate and relevant to the events of the day or the previous 24–48 hours.

Microsoft Copilot

I liked the choice. Copilot For headlines and sources, although its range of topics and number of headlines was lacking compared to Perplexity. However, each headline was up-to-date, covering either today’s news or the previous 24–48 hours. Check out the best AI search engines and tools you can use to search the web.

On the other hand, his descriptions were too short and to the point, which doesn't make for a straightforward way to read the news. Sure, it's useful as a starting point, but a one-line summary of a potentially complex news story doesn't do it justice.

It did a little worse when checking tech headlines, covering stories from the past week, and, interestingly, taking a headline from a podcast. So, not completely meaningless, but definitely not “breaking news” of the day.

Gemini

Of all the smart chatbots tested for daily news, it is perhaps the Gemini It's the best — although, as with other options, it really depends on your claim.

The general text prompt for today’s headlines doesn’t provide much information, typically returning four or five headlines with little analysis. But when asked to provide more information or provide more detailed news, it does so dutifully. Similarly, asking Gemini for general tech headlines gave me weeks’ worth of old news. But switching the prompt to “What are today’s tech headlines, please?” gave me the most accurate list of recent, breaking news of any chatbot I’ve tested.

The only downside is Gemini’s generic approach to sources. Instead of providing an individual link to each source for a news story, it throws up a list of sources at the end of its response, leaving you to search for the specific URL you need. Check out the best alternative ways to stay up to date without following the news.

Reasons Why You Shouldn't Rely on AI Chatbots for Your News

While AI can provide a convenient starting point for news information, AI chatbots also have significant limitations that need to be taken into account. Here are reasons why you shouldn’t rely solely on AI chatbots for your daily news updates.

1. lack of nuanceAI may struggle to provide comprehensive and accurate summaries of news events. Daily news updates lack context and the emotional weight of stories, while AI has difficulty capturing human emotions such as sadness, joy, or anger.
2. Inconsistency: Despite the improvement, you can’t expect the same output from the same prompt, even using the same AI model. You’ll likely miss headlines, see incorrect headlines, etc.
3. Limited coverageAs we saw above, the range of news provided by an AI chatbot is limited, even with an internet connection. An AI chatbot may also use questionable sources. As such, you are likely to miss niche topics (although it is fair to say that this can be overcome with proper text prompting).
4. hallucinationOne of the biggest problems is AI hallucination, where an AI model actively creates information and presents it as fact. As generative AI models get better, hallucinations have become a little less dangerous, but when it comes to the news, you want to make sure the information you’re reading is accurate.

Beyond that last point, I didn’t notice any AI hallucinations during testing. While both generative AI chatbots struggled to provide up-to-date news, none of the news they provided was hallucinations.

Overall, using AI to get my daily news was a mixed bag. Sure, the results were good once I tweaked the prompts, but by that point, I could just open my favorite news site, read the news, and move on with my day. Here are the best websites for cybersecurity and tech news.