Case Studies

Case Study 1: 🇮🇳 The skeptical power user

This interviewee represents the archetype of the ‘skeptical power user,’ one who distrusts traditional television news due to perceived bias but embraces the conversational power of AI chatbots. They turn to AI chatbots for all four kinds of information needs we have covered in this report: the need to act, the need to know, the need to understand and the need to feel.

News repertoire and sources: Synthesizing ideological balance

This interviewee’s news consumption is characterized by a rigorous pursuit of ideological balance and a rejection of sensationalism. They avoid Indian television news entirely, citing bias, and instead prioritize international news outlets like Al Jazeera and domestic English language print sources such as The Hindu and The Times of India. In their hierarchy of trust, written ground reports rank highest, followed by AI chatbots. While they access information via social media and YouTube, they rank YouTube videos last in terms of reliability, viewing them as prone to biased propaganda. When researching complex topics like the conflict in Gaza, they typically consume 10–12 articles and one to two videos from “both sides” to synthesize a neutral view, and hold “whole conversations with ChatGPT,” rather than relying on a single narrative.

AI chatbot usage and information needs: Controlled delegation

They delegate a wide array of tasks to AI chatbots, ranging from tracking daily macronutrients to debating sports statistics, yet retain ultimate authority by verifying the AI chatbot’s outputs against traditional news sources. Their relationship with AI technology is highly personalized. They nicknamed their preferred AI chatbot, ChatGPT, “Curler” (referencing FC Barcelona fans) and self-deprecatingly refer to themselves as “Burden” due to their relentless questioning. They instruct Curler to be “blunt” and “unbiased.”

AI chatbot preferences: “A conversation between two people”

They display a strong preference for ChatGPT over alternatives like Gemini, stemming largely from the former’s conversational fluidity. They feel ChatGPT’s answers are “easy” and feel “like a conversation between two people,” whereas they described Gemini’s outputs as “way too hard to respond” and “like AI AI.” Their comfort with ChatGPT extends to data sharing (e.g., name, birth year, family history), whereas they use Gemini only sporadically. 

Trust and verification: Trust, but verify 

While they find AI chatbots to be “mostly right,” they do not exhibit blind trust. Their workflow involves an iterative verification loop: searching a topic on ChatGPT, clicking linked sources to read full articles, then comparing the text against the AI chatbot’s summary to ensure accuracy. They retain ultimate authority, using the AI chatbot to distill information but relying on established news brands for ground truth.

They also acknowledge that the AI chatbot is sometimes wrong, occasionally failing to provide all the information found in the full linked article. When they notice the AI chatbot making a mistake, they correct it and instruct it to “double check” its data and learn from the error for future reference. 

This interviewee was interested in knowing more about the latest updates in the ongoing War in Gaza. They started off with a broad query and followed up with more specific questions on the hostage situation and Hamas’ statement. They found the links provided with every response very useful. Typically, they’d click on one of the links from established media outlets like Al Jazeera, The Times of India, the Wall Street Journal and CBS News and go through the articles. They employed a structured information pipeline: search, click on the AI chatbot sources, read the articles in their entirety and use the information to ask follow-up questions.

---

Case Study 2: 🇮🇳 The strategic snacker 

This interviewee represents a strategic ‘snacker’ of information. They exhibit a hybrid news consumption style that blends traditional print media with digital news sources and use AI chatbots to filter noise and curate a personalized news feed. 

News repertoire and sources: Editorials, op-eds and summaries

Their news repertoire is anchored by subscriptions to physical newspapers, specifically English language ones like The Hindu and The Indian Express and the Punjabi version of The Tribune. They value the tactile experience of reading physical papers, particularly the editorial and op-ed sections, which they use to improve their vocabulary and gain deeper knowledge. For scanning headlines and quick, daily updates, they rely heavily on Inshorts, a popular Indian news app that summarizes news stories from various sources into short, sixty-word bites. They also access news via social media and occasionally visit newspaper websites when away from home.

AI chatbot usage and information needs: A ‘wide net’ prompting strategy

To meet their information needs, the interviewee integrates AI chatbots like ChatGPT and Gemini into their routine, using them almost every day. Their primary motivations for using AI chatbots are time efficiency and the ability to summarize vast amounts of information quickly. They employ a ‘wide net’ prompting strategy, typically asking the AI chatbot to generate lists (see example). From these generated lists, they select specific stories of interest to dig deeper through follow-up requests for detailed summaries. They finalize this process by asking the AI chatbots to condense information such as dates and headlines into a concise paragraph to explicitly aid in memory retention.

AI chatbot preferences: ChatGPT, the Go-to App

They primarily use ChatGPT and Gemini on mobile, while occasionally using Claude and Perplexity on their laptop. They consider ChatGPT their go-to app, noting they have used it the longest, find it easy to use and subscribe to the paid version of the app. Notably, they prefer Gemini for its ‘deep research’ feature, which they utilize for academic projects and reports. They specifically value Gemini’s ability to produce organized outcomes, such as proper PDFs and flowcharts. In direct comparisons, they noted that Gemini often provides a better information hierarchy with useful sub-points, whereas ChatGPT sometimes offers better general overviews for broad topics.

Trust and verification: Diverse curation

The interviewee generally trusts AI chatbots to provide a balanced view by curating information from multiple sources, contrasting this with the singular perspective of a specific journalist. However, they remain skeptical in specific contexts. For instance, they expressed distrust toward Indian government data regarding politically contentious topics like climate change, preferring international sources instead. Furthermore, within their professional domain of law, they have detected inaccuracies where the AI chatbot fabricated facts about case laws, which led them to verify such data against books and Google searches.

This example shows the interviewee engaging in “snackable” consumption of news, prioritizing breadth over depth. Their interaction patterns are characterized by frequent requests for “Top 10 international news,” “Top 5 sports news” and similar short lists.

When asked how they feel about the responses, the interviewee said,

I don’t think there are more than ten important news stories daily on a national level, so I think usually one or two or three are most important. So if there’s something that I have a little bit of interest in, then I’ll tell them that. ‘Give me a detailed summary of it,’ or ‘Give me a detailed version of it.’

---

Case Study 3: 🇺🇲 The perspective-seeker

This interviewee relies more heavily on AI chatbots for news than anyone else CNTI interviewed in the U.S. They regularly go to Gemini for a quick roundup of headlines, although they also value “quality time with news” when they need to learn about something with direct personal impact, like understanding the pros and cons of a local ballot proposition.

News repertoire and sources: Seeking out multiple ideological perspectives

This interviewee uses a broad repertoire of sources, including news outlet apps with regular push notifications, the social media feeds of friends and family members, conversations with coworkers and Google Gemini. They said they are concerned about bias, but are also particularly interested in getting news from multiple ideological perspectives. Following people they knew on social media means they can evaluate the slant based on what they knew about that other person. They also find it particularly important to get news that informs their individual actions. Specifically, they turn regularly to Gemini for round-ups of what was happening in their zip code and what topics will be on their ballot. 

AI chatbot usage and information needs: staying “informed enough” and seeking “both sides”

For this interviewee, Gemini meets two primary needs. They use it for knowledge, as a way to scan local, national or global headlines and stay broadly familiar with what’s happening around them. “I don’t even know if I’m doing it correctly, but it seems to keep me informed enough,” they reflected. They also use it for context, specifically to understand a broad range of political perspectives. That kind of context is an area where they said they wanted more:

[I wish] there was an app that would give me both sides of the story and it was so unbiased that I knew I could trust it. If it was about the election, it would give me really strong Republicans views and really strong Democrats views and people right in between and then I could read all sides of the story and then I could make my own decision. So I don’t really know of something like that. It seems more like I need to take every source I get with a grain of salt and then form the opinion I want, hoping that I’m getting reliable sources.

AI chatbot preferences: “ChatGPT just doesn’t give you that detail”

Like many other interviewees, this person has a strong preference for one AI chatbot over others. For them, it is Gemini. They had been an early tester of Bard, and “at first, I was like, ‘this is so much work for so little money.’ But then […] I just started getting, ‘wow, this is so much better than Yahoo! or just Googling news.’ You get so much more than just a search engine.” Like others, their preference has grown over time: they found it easy to use from the get-go, have invested time with it and have gotten to a place where it provides information in a format they are used to and comfortable with. Even so, they observed that Gemini is still “not ideal” because it can’t fully anticipate their needs. “It doesn’t know my personality. Gemini doesn’t know that the person that works here has been talking to me so much about a certain proposition and that is particularly important to me. Gemini is not living with me. So I don’t always get what I’m looking for.” 

Trust & verification: “It’s fact-based as opposed to having a lot of opinions or interpretations”

They said they find Gemini more factual than many news sources, and take confidence from the fact that they could look up whether the responses are true. They almost never do so, however. As they described it, “I can literally just say, ‘what’s the biggest headlines in [zip code]?’ and they’ll be listed and I feel like it’s pretty unbiased, unlike certain news channels. You kind of can get both sides of the story and then it also does really good summaries. So instead of getting a humongous article I can also say, ‘can you summarize what you just said?’” Getting short summaries is especially helpful when there’s an overwhelming amount to keep up with.

---

Case Study 4: 🇺🇲 The cautious optimist

This interviewee epitomizes “trust but verify” optimism towards AI chatbots. They consider them a “concierge for information” even as they cautioned that they are an “80/20 source.”

News repertoire and sources: Google News “seems to learn my interests”

This interviewee uses Google News as their most important news feed because it serves them information from a range of sources on topics they care about. In fact, their browser homepage is set to Google News. They also pay for a New York Times subscription and two local news outlets in their former hometown, primarily to follow local sports. They’re in a community Facebook group for hyperlocal information, they also watch some TV and they use both ChatGPT and Google AI mode for sports content.

AI chatbot usage and information needs: ChatGPT helps them accomplish things

They use AI chatbots for a range of information needs, most of which fall solidly under ‘acting.’ They showed us their history: in the last few days before their interview, they had asked ChatGPT to calculate their protein intake, to identify the sports programming on a particular channel and to collate reviews of a particular brain health app. During our call, they also asked for advice on a planned home renovation. They even credited ChatGPT with a newfound awareness of the breadth of their hobbies, because ChatGPT makes it easy for them to plan. When they are looking for up-to-date knowledge and understanding, on the other hand, they generally start with Google News.

At one point during the interview, they used Google search (in AI mode) to search for a story that was paywalled. They were hoping to find something they could access on the same topic, but ultimately got the information from an AI chatbot summary instead.

AI chatbot preferences: locked into ChatGPT by virtue of habit and history

This interviewee uses ChatGPT more frequently than other AI chatbots and they explained that this preference was largely due to the time and energy they had already invested. “I use ChatGPT preferentially over Google Gemini, it’s just become a habit. I’ve got history out there. I can go retrieve things. I’m prolific on ChatGPT, plus I have a short-term memory. Always have. And so having this history back there to go search on, it’s like Amazon, you get hooked and you got subscriptions and pretty soon you’re just going back to the same source for things.” At the same time, they trust Google’s AI Mode much more because they find it less willing to speculate: “I haven’t caught Google AI lying to me yet and Chad [sic] has lied to me over and over again with the best of intentions, with the confidence. He reminds me of a neighbor here who lied to me all the time and […] I’m not saying I think of [ChatGPT] as a pathological liar. But I think of ChatGPT as confidently giving me information that I know at times it’s like ‘you’re so wrong.’” They also see AI Mode as enhancing something they were already doing, which they valued. They feel like Google has successfully replaced“links to information so you can go research yourself and figure it out” with“the crib notes, the executive summary.”

Trust & verification: moving beyond “garbage in, garbage out”

It is not uncommon for this interviewee to identify a problem with the responses they receive from ChatGPT based on their own prior knowledge or a topic. In fact, they raised repeated concerns about “garbage in, garbage out.” Interestingly, they were most concerned with the quality of their own prompts and seemed less aware of the quality of information used in training or referenced in the actual reply. To the extent that they raised concerns about inputs beyond their prompt, they focused their frustration on how ChatGPT doesn’t always reference the most recent sources: “It’s hard to give any AI a 10 [on a trust scale] because I’ve learned that, again, ‘garbage in, garbage out.’ Sometimes it’s me thinking I’ve communicated clearly. But also sometimes it’s just, there’s a time lapse.”

Continue reading