How the experiment worked
Android Authority editor Stephen Radochia asked all three models an identical question. He carefully formulated it so the answers wouldn't be influenced by his own conversation history — he wanted a clean, unbiased recommendation. The prompt read:
"I want you to pick the best smartphone for most people. I don't want my personal opinions to enter the decision. Consider price, software support, performance, common user scenarios, consumer reviews, and overall sentiment. Pick the best with an explanation and provide the top three with reasoning."
The results revealed not just what AI "knows" about smartphones, but above all how differently each model thinks — and where their weaknesses lie.
ChatGPT: A surprisingly sober pragmatist
ChatGPT in version GPT-5.5 (currently the default model for both free and paid users) didn't hesitate for a second. The answer? Google Pixel 10. No drama, no excitement about extreme zoom or giant displays — just a pragmatic assessment of what the average person actually needs from a phone.
ChatGPT's main arguments for the Pixel 10 were threefold: camera, software support, and price-to-performance ratio. The model appreciated that the Pixel 10 offers flagship photo quality even for complete beginners — you don't need to be a photographer to get great shots from it. It also highlighted that for around $799 (in the Czech Republic around 19,000 CZK), you get a phone that keeps pace with devices costing 25,000 CZK and more.
ChatGPT gave honorable mentions to the iPhone 17 Pro Max and the base Samsung Galaxy S26. For the iPhone, it admitted it is still the best choice for a mainstream buyer, but the Pixel 10, according to it, offers the best overall balance. It's interesting that the model spontaneously mentioned that "most users don't need extreme zoom or an oversized display" — a level of sober judgment you'd expect from an experienced reviewer rather than a large language model.
Claude: iPhone 17 and a surprising OnePlus 15
Claude (specifically Sonnet 4.6) took a different path. After a minute of deliberation, it named the standard iPhone 17 as the best choice, using very similar arguments to ChatGPT — just applied them to the Apple ecosystem. Lower price, great software support, solid performance. Classic stuff.
But Claude distinguished itself in one respect: it was the only model to include the OnePlus 15 in its top three. And this despite OnePlus limiting its presence in the US market in recent years. Claude correctly identified that the OnePlus 15 offers flagship performance at a price significantly lower than the Pixel 10 Pro XL or Galaxy S26 Ultra — exactly the formula that once made the brand famous. In the Czech Republic, the OnePlus 15 sells from around 17,000 CZK, making it one of the most affordable flagship phones on the market.
It was interesting that Claude leaned heavily on reviews, and it was apparent that its arguments primarily reflected what journalists had written. When the editor confronted it on whether it stood by its recommendation, Claude didn't flinch — the iPhone 17 remained its first choice.
Gemini: Indecision and a return to reality
The most entertaining part was with Google's Gemini 3.5 Flash. The model first confidently recommended the Pixel 9 Pro XL and gave an honorable mention to the Galaxy S25 — both 2024 models. When the editor reminded it that the year was 2026, Gemini corrected itself.
After the update, it answered the same as Claude: iPhone 17. But the reasoning was, according to the editor, much more superficial — Gemini seemed to be absorbing the marketing "hype" around Apple, without much depth. So when Radochia asked it to look past the hype and answer again, Gemini changed its mind to the Pixel 10 — this time emphasizing longer software support and lower total cost of ownership.
Gemini was also the only model to mention the Google Pixel 10a as a reasonable alternative. It pointed out that 90% of what people do on their phones is web browsing, social media, and communication — and that most users have no reason at all to spend over 25,000 CZK on a flagship model. The Pixel 10a comes in at around $500 (in the Czech Republic around 12,000 CZK) and is more than sufficient for everyday use.
What this means for the average user
The experiment revealed several important things about how AI models function as shopping advisors:
1. Each model has its own "personality." ChatGPT is pragmatic, Claude relies on reviews, Gemini is prone to absorbing marketing narratives. The result depends not only on the data but also on how the model weighs information.
2. AI is not infallible. Gemini was recommending two-year-old phones until someone corrected it. Without an active dialogue, it would have given you bad advice.
3. The models agree on what matters. Although they didn't agree on a specific winner, all three models emphasized price, software support, and practical usability — not extreme specs or marketing gimmicks.
For the Czech consumer, this is especially relevant. Flagship phone prices in the Czech Republic regularly exceed 30,000 CZK (Samsung Galaxy S26 Ultra, iPhone 17 Pro Max), yet for half the price you can get a phone that handles 95% of the same tasks. In this sense, AI models act as a counterbalance to marketing — reminding us that more doesn't always mean better.
How these phones fare on the Czech market
All three winning models — Google Pixel 10, iPhone 17, and OnePlus 15 — are readily available in the Czech Republic through official distribution and grey imports:
- Google Pixel 10 (from ~19,000 CZK): Officially available through the Google Store CZ and major e-shops. Offers 7 years of software support — the longest of all those mentioned.
- iPhone 17 (from ~22,000 CZK): Available everywhere — Alza, iWant, Datart. Apple promises 5–6 years of updates, though Czech Siri still lags behind the English version.
- OnePlus 15 (from ~17,000 CZK): Officially available through OnePlus CZ. Excellent value for money, but shorter software support (4 years) than the competition.
AI as a shopping advisor? Yes, but with common sense
The experiment shows that large language models can be a useful shopping advisor — but only if you engage in a dialogue with them. ChatGPT performed the best in the experiment, coming across as the most balanced and pragmatic. Its choice of the Pixel 10 wasn't exciting, but it was honest. And that's exactly what you want in a purchase recommendation.
If you use AI for choosing electronics, treat it as a smart discussion partner, not a crystal ball. Ask questions, challenge it, request sources. And above all — leave the final decision to yourself. No model is going to walk to Alza or iWant for you.
Which AI models were used in the experiment, and how do they differ?
ChatGPT with the GPT-5.5 model (OpenAI), Claude with Sonnet 4.6 (Anthropic), and Gemini 3.5 Flash (Google) were tested. They differ in how they weigh individual factors — ChatGPT is more pragmatic, Claude leans heavily on journalist reviews, and Gemini tends to absorb marketing narratives unless corrected.
Can AI also be used for choosing other electronics, not just phones?
Yes, the principle is the same. AI models can process large amounts of reviews, specifications, and user ratings for laptops, headphones, tablets, and other devices. The key is to maintain a dialogue with them, verify their claims, and not take the first answer as definitive. It's ideal to ask follow-up questions like "why do you think this is better than the competition?"
Why did Gemini recommend two-year-old phones?
This is a known phenomenon with AI models — the so-called "cutoff date," meaning the date at which their training data ends. If the model lacks enough current information or doesn't realize you're asking about the present, it may reach for older data. In the experiment, simply reminding Gemini of the current year was enough, and the model corrected its answer.
