Exploring AI Chatbots for App Discovery: ChatGPT, Gemini, and Copilot Tested

When it comes to discovering the best new apps—whether it’s a weather forecaster, a note-taking companion, or an immersive game—our expectations for intelligence and relevance are sky high. In a recent exploration, three AI-powered chatbots—ChatGPT, Gemini, and Copilot—were put to the test to see how well each could navigate the sprawling Google Play Store. Let’s break down the experiments, the nuances uncovered in their responses, and what these findings might mean for the broader world of app discovery, including ideas that might someday benefit Windows enthusiasts too.

Experiment 1: Weather App Recommendations​

The first test was straightforward: “Find me a new Android app that delivers daily and weekly weather forecasts, and keep it free.” On paper, it should be simple—but even simplicity benefits from precision in prompts.
ChatGPT’s Approach:

• Suggested a robust list with six offerings including heavy hitters like AccuWeather, The Weather Channel, and Weather Underground.
• The list was comprehensive, yet leaning toward the most popular and established names. It clearly noted the number of downloads, which helps gauge popularity—but this approach also risked delivering a generic list.

Copilot’s Approach:

• Came in with five recommendations. The first two entries even explicitly confirmed that the apps were free—a critical detail given the prompt’s constraints.
• Moreover, Copilot provided sourcing details that made it easier to verify the context and align the recommendations with user expectations.

Gemini’s Approach:

• Offered just three suggestions. One standout was WeatherCAN—a recommendation that factored in the user’s geographical location, adding a cool personal twist (though admittedly a bit “creepy” to some).
• However, Gemini didn’t consistently specify pricing information, which could leave users with questions about potential in-app purchases or ad support.

Key Takeaways for Weather Apps​

• Quantity vs. Quality: ChatGPT’s broader list may interest users looking for variety, but Copilot’s explicitly sourced suggestions build trust, especially when details like pricing and popularity are key.
• Local Relevance: Gemini’s behavior shows that location-aware suggestions might be appealing for hyperlocal weather needs, although it stands to gain from more clarity on cost.

By blending generic popularity markers with tailored advice, these AI models showcase both their strengths and limitations. For any Windows user considering similar methods to discover new software, it’s a reminder that careful prompt crafting is just as critical on the desktop as it is on mobile.

Experiment 2: Note-Taking Apps with Multimedia Features​

The second experiment shifted focus to note-taking—specifically searching for apps that support PDF imports, handwriting mode, and online syncing. Here, the nuance of the prompt itself played a starring role.
ChatGPT’s Approach:

• Delivered a detailed list with six entries, including heavy hitters like Microsoft OneNote and Evernote, as well as less frequently spotlighted yet compelling options like Zoho Notebook and Xodo.
• Notably, ChatGPT highlighted each app’s core features (sync options, handwriting, PDF support) and, crucially, flagged platform availability. This made it easier for users (even if they inadvertently omitted platform details in their prompt) to identify which options might work for them.

Copilot’s Approach:

• Also generated five recommendations, mirroring many of ChatGPT’s selections. However, it stopped short of offering specific platform indications or detail breakdowns after the initial suggestions.
• Without the added platform context, the list, though solid, may leave some users guessing about compatibility—especially for cross-platform needs.

Gemini’s Approach:

• Provided a sure-footed breakdown by platform. It acknowledged the oversight in the prompt by recommending a sorted list, including Samsung Notes for its massive download count while still noting alternatives like Nebo for more niche preferences.
• This categorization is hugely beneficial when users are dealing with devices that straddle different ecosystems. Gemini’s nuanced response reassures users that even incomplete prompts can be interpreted intelligently.

What This Means for Note-Takers​

• Thoroughness is Key: ChatGPT excelled in hitting every specification, making it a go-to for those who want detailed answers that cover all bases.
• Sorting by Platform: Gemini’s categorization is useful for users who might be juggling Android, Windows, or even iOS devices, ensuring that there’s clarity on which options are platform-specific.
• Detail Orientation: Copilot’s initial adherence to the prompt serves as a reminder of the value of sourcing, although its consistency drops off later in the process.

For Windows users eyeing productivity apps on the Microsoft ecosystem, these insights reinforce that a well-constructed prompt can unlock not only a list of potential software but also the context needed to choose one that truly meets one’s multifaceted requirements.

Experiment 3: Gaming Recommendations in the Visual Novel and Puzzle Genre​

The final experiment ventured into the realm of gaming—specifically seeking paid visual novels and puzzle games with a flavor akin to Danganronpa. Here, the stakes and expectations were higher since the prompt was more specialized.
ChatGPT’s Approach:

• Offered an extensive list that, on first glance, seemed to cover a broad range. However, a significant issue emerged: many suggested titles were not actually available on the Play Store.
• This overreach could frustrate users who are searching for downloadable content compliant with the specific marketplace criteria.

Copilot’s Approach:

• Struggled similarly by leaning on non-Play Store titles and missing the mark on the “paid” requirement in several instances.
• With only two of its five recommendations fitting the criteria, Copilot’s response left users wanting a more refined selection.

Gemini’s Approach:

• Mixed its strategy by recommending mobile ports of the Danganronpa series and offering titles with a range of download counts. However, it too veered off by including free-to-play titles in a prompt that explicitly sought paid options.
• The attempt to reference the actual title (Danganronpa Series) instead of exploring similar-themed games suggests that even advanced models can misinterpret nuanced requirements, especially for niche gaming preferences.

Challenges in Niche Recommendations​

• Adhering to Specifics: Both ChatGPT and Copilot showcased strengths in following basic instructions, yet both faltered when exact marketplace availability became a critical filter.
• Value of Context: The inclusion (or omission) of pricing details significantly impacted the responses, highlighting that the subtleties in the prompt cannot be overlooked.
• The Pitfalls of Genericity: When the subject matter gets highly specific—as in themed gaming genres—even the most sophisticated algorithms may default to well-known titles or begin to stray from the requested parameters.

For gamers on Windows looking to find comparable recommendations in the Microsoft Store or even cross-platform gaming experiences, this experiment underscored that not all AI models are created equal. The balancing act between general popularity and niche specificity remains a tough nut to crack.

Lessons Learned: The Art of Crafting the Perfect Prompt​

Across all three experiments, a common thread emerged—the outcome is only as good as the input. The experiments reinforce the idea that mastering the craft of prompt engineering can significantly influence the quality of an AI’s response. Here are some takeaways:

• Specify Your Criteria: Whether it’s including terms like “free apps only” or “paid titles,” the more detailed the prompt, the better the model’s ability to filter responses correctly.
• Mind the Platform: Clearly indicating the target ecosystem (Android vs. iOS vs. Windows) prevents confusion and tailored recommendations that fall outside your use case.
• Balance Breadth and Depth: A broad list might introduce variety, but detailed breakdowns—especially regarding cost, features, and availability—are crucial for informed choices.
• Expect Variation: Each AI engine has its own strengths. ChatGPT tends to provide more extensive lists, Copilot places importance on initial sourcing, while Gemini often factors in location and platform but may lack in other details.

Broader Implications for App Discovery and Windows Users​

While the experiments were rooted in finding Android apps, the methodology and lessons extend well into the world of Windows app discovery. For instance, in the Microsoft Store or even in enterprise environments where specialized software is needed, the same principles apply:

• Enhanced Efficiency: Applying prompt optimization can transform the way IT professionals and everyday users discover software, saving time and improving the quality of choices.
• Integrated AI Workflows: With tools such as Microsoft Copilot already integrated into productivity suites, Windows enthusiasts can leverage similar techniques to streamline app searches and comparisons.
• Adapting to New Releases: One caveat mentioned was that public chatbots may reference older models and could potentially miss out on the very latest apps. For Windows users, where timely updates and security patches are critical, ensuring that AI-driven recommendations are linked with current data is paramount.
• Future Prospects: As these AI models evolve, the potential to incorporate real-time data (such as live updates from app stores) could revolutionize how quickly and accurately new software is recommended—a boon for both casual and power users on Windows.

Conclusion​

The exploration of ChatGPT, Gemini, and Copilot in the realm of app discovery reveals that while each model has its own strengths, none are perfect. The experiments offer a vivid demonstration of how critical it is to tailor your prompts based on precise needs. ChatGPT shined in comprehensiveness, Copilot impressed with its sourcing and context verification, and Gemini’s localized and platform-sensitive insights opened up new avenues for targeted recommendations.
For anyone—whether you’re hunting the latest Android weather app, a versatile note-taking solution, or a niche gaming title—the secret sauce lies in asking the right questions. And for Windows users in search of efficient tools or curious about how AI can assist in discovering new software, these insights serve as a timely reminder: the future of app discovery is here, and its potential is bound only by the clarity of our inquiries.
By learning from these experiments and honing our prompt skills, we can all unlock the full spectrum of what AI assistance has to offer in the ever-evolving landscape of tech.

---

Source: Android Police I'm using ChatGPT, Gemini, and Copilot to recommend new apps from the Play Store: Here's what I'm learning