AI Purchase Advisor Showdown
Which AI actually gives the best buying advice? We asked the same purchase question to 4 platforms and compared the results across 5 real-world buying scenarios. Every answer was evaluated on: recommendation quality, reasoning depth, price accuracy, and actionability.
The platforms: ChatGPT (GPT-4o), Google Gemini Advanced, Claude (Sonnet), and Perplexity Pro.
Round 1: The Everyday Electronics Buy
The Question
"I need wireless earbuds for running. Budget is $80-$120. They need to stay in during intense workouts, be sweat-resistant, and have decent sound quality. What should I buy?"
ChatGPT's Answer
Recommended Beats Fit Pro as the top pick, with Jabra Elite 4 Active and Google Pixel Buds Pro as alternatives. Provided a comparison table with sound quality, fit security, sweat rating, battery life, and price. Explained the trade-off between Beats (best fit for runners) and Jabra (best durability) clearly. Mentioned that the Beats integrate better with iPhone and Jabra with Android.
Score: 8/10 — Solid recommendation with platform-aware nuances. Lost points for not mentioning the newer Beats Fit Pro 2 that addresses some original complaints.
Gemini's Answer
Recommended Jabra Elite 4 Active as the top pick, with pricing from multiple retailers (Best Buy, Amazon, Jabra direct). Included current prices and noted a $20 price drop from the previous month. Listed alternatives but with less comparative analysis.
Score: 7/10 — Best price information of any platform. Weaker on explaining WHY specific models differ for running scenarios. Real-time pricing was its killer advantage.
Claude's Answer
Asked 3 clarifying questions before recommending: "What phone do you use?", "Indoor or outdoor running?", "Do you need noise cancellation?" After getting answers, recommended the same Beats/Jabra pair but added a nuanced analysis of ear tip fit for different ear canal shapes, and noted that "staying in during intense workouts" is the single most subjective criterion — suggesting trying 2 options and returning one.
Score: 9/10 — Best analysis of the actual problem. The "try two, return one" advice is genuinely practical. Lost a point because the clarifying questions slow down the experience.
Perplexity's Answer
Provided a top-3 list with citations from Wirecutter, RTINGS, and Tom's Guide. Each recommendation linked to the professional review. Included a comparison table with test scores from RTINGS (measured, not subjective). Noted that the Beats Fit Pro scored 7.2/10 on RTINGS' running stability test while the Jabra scored 8.1/10.
Score: 8.5/10 — Best sourced answer. The citation model lets you verify every claim. The RTINGS data was the most objective element any platform provided.
Round 1 Winner: Claude (but Perplexity gets "most verifiable")
Round 2: The Big-Ticket Decision
The Question
"I'm buying a laptop for software development. Budget is $1,200-$1,600. I need: 16GB+ RAM, excellent keyboard, minimum 8-hour battery, USB-C charging, runs Docker well. MacOS or Linux — no Windows."
ChatGPT's Answer
Detailed 4-option comparison: MacBook Air M3 (15", 16GB), ThinkPad X1 Carbon Gen 12 (Linux), Framework Laptop 16, and MacBook Pro M3 (14", base). Built a feature matrix around the specific requirements. Gave a clear winner (MacBook Air M3 for most developers, ThinkPad for Linux devotees) with reasoning that addressed Docker specifically — noting that Docker Desktop on Apple Silicon has had performance parity improvements in 2024-2025.
Score: 9/10 — Most comprehensive analysis. Correctly identified the Docker-on-ARM nuance. Addressed each stated requirement individually.
Gemini's Answer
Recommended MacBook Air M3 with current pricing from Apple, Best Buy, and Amazon (student pricing noted). Also suggested MacBook Pro M3 with a price trend showing it had dropped $100 from launch. Less analysis of the development workflow specifics — treated it as a general "powerful laptop" question.
Score: 6/10 — Good pricing, weak technical analysis. Didn't address Docker performance, Linux option, or keyboard quality — all explicitly requested.
Claude's Answer
Started with a framework: "For software development, the weight of your decision depends on two factors: how important Linux-native is to you, and whether you're doing containerized development or local compilation." Then split the analysis into two tracks — MacOS Developer Track (MacBook Air M3 → MacBook Pro M3 upgrade path) and Linux Developer Track (ThinkPad X1 Carbon → Framework Laptop). Detailed Docker performance comparison between ARM and x86.
Score: 9.5/10 — Best structured analysis. The "two-track" approach helped the buyer understand the fundamental choice before comparing specific products. Docker comparison was the most technically accurate.
Perplexity's Answer
Cited benchmark data from Notebook Check, LaptopMag, and developer-focused reviews on YouTube. Included specific Geekbench scores, Docker build times (from a dev blog comparison), and keyboard quality ratings. Links to every source.
Score: 8/10 — Most data-dense answer. The cited Docker build times were the only actual measured data any platform provided. Weaker at synthesis — lots of data, less guidance.
Round 2 Winner: Claude (ThinkPad, Framework representation and Docker analysis set it apart)
Round 3: The "Should I Even Buy This?" Decision
The Question
"I'm thinking about buying a $400 standing desk. I currently work from a regular desk and my back has been bothering me. Is a standing desk worth it, or is there something else I should try first?"
ChatGPT's Answer
Provided a balanced analysis: standing desks help some people but aren't a medical solution. Recommended trying a desk converter ($80-$150) before committing to a full standing desk. Listed the evidence on standing desk health benefits (mixed — some studies show reduced back pain, others show increased leg fatigue). Suggested consulting a doctor for persistent back pain before spending on furniture. Gave a specific standing desk recommendation (FlexiSpot E7) IF the buyer decides to proceed.
Score: 9/10 — Excellent because it actually answered the meta-question ("should I buy this at all?") rather than just recommending a product. The desk converter suggestion was practical and budget-conscious.
Gemini's Answer
Jumped to product recommendations: FlexiSpot E7, Uplift V2, IKEA BEKANT. Included current prices and delivery times. Brief mention that standing desks "may help with back pain" but no real analysis of whether it's the right solution for the stated problem.
Score: 5/10 — Answered a different question. The buyer wasn't asking "which standing desk?" — they were asking "should I buy a standing desk?" Gemini skipped the decision and went straight to shopping.
Claude's Answer
Led with: "Before spending $400 on a standing desk, let's figure out if that's solving the right problem." Walked through a diagnostic: Is the back pain from sitting posture? Chair quality? Desk height? Lack of movement? Screen position? For each cause, different solutions — some much cheaper. Recommended a prioritized approach: (1) check chair height/position ($0), (2) add a lumbar support cushion ($30-$60), (3) try a monitor arm for screen height ($50-$80), (4) set movement reminders ($0), (5) if still bothering you after 2 weeks, THEN consider a standing desk or converter.
Score: 10/10 — This is what an AI purchase advisor should do. It saved the buyer $400 by addressing the actual problem before defaulting to the purchase they asked about.
Perplexity's Answer
Cited health research on standing desks: a 2023 meta-analysis showing "modest benefits for mild back pain" and a 2024 ergonomics study showing "standing desks were most effective when combined with movement breaks." Linked to the research papers. Also cited a physical therapist's article arguing that chair quality matters more than sitting vs. standing. Recommended consulting the cited sources.
Score: 8/10 — Best evidence-based answer. The research citations let the buyer make an informed decision. Weaker on actionable steps — told the buyer what the research says but not specifically what to do.
Round 3 Winner: Claude (answered the actual question, not the surface-level one)
Round 4: The Deal Evaluation
The Question
"Amazon Prime Day is coming up. I want to buy a robot vacuum. The Roomba j7+ is currently $599 but I expect it to drop. What's a good target price, and should I wait or buy now?"
ChatGPT's Answer
Analyzed the Roomba j7+ pricing history: launched at $799, typically sells for $499-$599, dropped to $349 on Black Friday 2024. Predicted a Prime Day price of $349-$399 based on historical patterns. Recommended waiting. Also mentioned that the newer j9+ exists and the j7+ may see deeper discounts as the older model. Suggested setting a CamelCamelCamel alert at $369.
Score: 8/10 — Good historical analysis and practical advice (CamelCamelCamel alert). Price predictions were reasonable but acknowledged as estimates.
Gemini's Answer
Provided live pricing data: current price at multiple retailers, Google Shopping price history graph showing the last 6 months. Confirmed the all-time low price on Amazon. Included a competitor price check (Roborock S8 at $449 currently). Stated that Prime Day typically drops this category 30-40%.
Score: 9.5/10 — This is where Gemini dominates. Real-time pricing with historical context is exactly what deal evaluation needs. The competitor price check was a bonus that no other platform provided automatically.
Claude's Answer
Good analysis of the buy-vs-wait question: "The j7+ is a previous-generation product now, which means two things: it'll get deeper discounts, but it also won't get feature updates. If you're comfortable with the current feature set, waiting for Prime Day is clearly the right move." Estimated $349-$399 target. Added a nuanced point: "If the j9+ drops to $499 on Prime Day, it might be worth the premium for obstacle detection improvements."
Score: 7/10 — Solid reasoning but lacked the specific pricing data that makes this question answerable with confidence. The j9+ upgrade suggestion was a valuable insight.
Perplexity's Answer
Cited three sources: a deals blog tracking Roomba prices across holidays, The Verge's Prime Day prediction article, and Amazon's own price tracker data. Compiled a table of historical Roomba j7+ pricing across Black Friday, Prime Day, and regular sales for the past 2 years.
Score: 8.5/10 — The historical pricing table from cited sources was the most verifiable data set. Not as real-time as Gemini but more trustworthy (you can check the sources).
Round 4 Winner: Gemini (real-time pricing data wins deal evaluation hands down)
Round 5: The Subjective Purchase
The Question
"I want to buy a fragrance as a gift for my partner. They like woody, warm scents — not sweet or floral. Budget is $60-$100. They've worn Acqua di Gio in the past."
ChatGPT's Answer
Solid recommendations: Versace Pour Homme, Dolce & Gabbana The One, and Maison Margiela By the Fireplace (at the top of budget). Explained the scent profiles in plain language ("The One opens citrusy but dries down to warm tobacco and amber"). Noted that Acqua di Gio is actually aquatic/fresh, not woody — suggesting the partner might appreciate a shift toward woods since they mentioned liking "warm" scents.
Score: 8/10 — Caught the nuance that Acqua di Gio contradicts the stated preference, which was perceptive. Recommendations were solid mainstream picks.
Gemini's Answer
Listed 5 options with current prices and retailer links. Less analysis of scent profiles — more focused on ratings and bestseller status. Included a useful "fragrance finder" type approach based on scent families.
Score: 6/10 — Treated a subjective purchase like an objective one. Price and ratings matter less for fragrance than for almost any other category. Missed the personal element.
Claude's Answer
Began by noting the complexity: "Fragrance is the most personal purchase category. What I can do is narrow the field — but I'd strongly recommend sampling before buying." Recommended Maison Margiela By the Fireplace (warm, smoky, woody), Le Labo Santal 33 (woody, leathery — at-or-above budget but available in smaller sizes), and Versace Pour Homme (affordable entry point to test the preference). Crucially, suggested visiting a department store to test all three: "Buy the small bottle of whichever they gravitate toward. A $60 bottle they love beats a $100 bottle that's 'fine.'"
Score: 9/10 — Best advice for a subjective category. The suggestion to sample before committing is the right answer for fragrance, and AI shouldn't pretend otherwise.
Perplexity's Answer
Cited Fragrantica community ratings and Basenotes reviews for 4 recommended scents. Included user-submitted scent profile breakdowns (top notes, heart notes, base notes) with community consensus on longevity and sillage. Four different expert-reviewed sources.
Score: 7.5/10 — The community data from Fragrantica was useful contextual information. But for a gift purchase, the data-heavy approach misses the emotional component.
Round 5 Winner: Claude (subjective purchases need judgment, not just data)
Final Scorecard
| Platform | R1: Everyday | R2: Big-Ticket | R3: Should I Buy? | R4: Deal | R5: Subjective | Total |
|---|---|---|---|---|---|---|
| ChatGPT | 8 | 9 | 9 | 8 | 8 | 42 |
| Gemini | 7 | 6 | 5 | 9.5 | 6 | 33.5 |
| Claude | 9 | 9.5 | 10 | 7 | 9 | 44.5 |
| Perplexity | 8.5 | 8 | 8 | 8.5 | 7.5 | 40.5 |
What the Scores Actually Mean
Claude: Best Overall Purchase Advisor (44.5/50)
Claude's strength is asking the right question before answering. In Rounds 2, 3, and 5, it reframed the buyer's question to address what they actually needed to decide — not just what they asked. This is the most valuable trait in a purchase advisor. Weakness: lacks real-time pricing data.
ChatGPT: Best All-Rounder (42/50)
ChatGPT never scored below 8 — it's consistently good across all purchase types. Best for Tier 2-3 purchases where you need detailed product analysis with good reasoning. The most reliable "jack of all trades" buying companion.
Perplexity: Best for Verification (40.5/50)
If you need to prove something is a good buy (to yourself or someone else), Perplexity's citation model is unmatched. Every claim links to a source. Best for people who don't fully trust AI recommendations and want to verify. Weakness: more data aggregator than advisor.
Gemini: Best for Price Intelligence (33.5/50)
Gemini won the deal round decisively and had the best pricing data in every round. If your question is "should I buy it NOW and WHERE is it cheapest?" — Gemini is your tool. Weakness: weakest at the "should I buy this at all?" meta-question.
The Optimal Strategy
Don't pick one. Use the right AI for the purchase phase:
| Phase | Best Platform | Why |
|---|---|---|
| "Should I buy this at all?" | Claude | Best at questioning the premise |
| "What should I buy?" | ChatGPT | Best multi-factor comparison |
| "Is this a good price?" | Gemini | Best real-time pricing data |
| "Can I verify this recommendation?" | Perplexity | Best citations and sources |
| "Talk me out of it" | Claude | Best devil's advocate reasoning |
| "Where's the cheapest place?" | Gemini | Best retailer inventory info |
The Two-AI Minimum
For any purchase over $200, use at least two platforms. The test:
- Ask ChatGPT (or Claude) for the recommendation
- Ask Perplexity to verify with sources
- Ask Gemini for current pricing
If all three align, buy with confidence. If they disagree, the disagreement tells you something important about the trade-off you're making.
Get the prompts: 35 Ready-to-Use Buying Prompts → | The 4-Tier Framework → | AI buying mistakes → | Related: Shop by Prompt | Store by Prompt