Best AI for Searching Online: Tested Across 63 Queries
Overview with some in-depth comparative analysis; limited technical depth on how the tools work internally
Easy to understand, conversational, with a personal-experience tone and a promotional undercurrent for the author's newsletter
0 0 1
Ruben Hassid tests seven AI tools (ChatGPT, Perplexity, Grok, DeepSeek, Claude, Gemini, Kimi) across nine search queries to identify which performs best for online searching. He finds Perplexity the clear overall winner for search, with strengths in source aggregation, YouTube summarization, Reddit mining, and formatting. He compares standard search against Deep Research modes (longer, more sources, but much slower), recommends specific setups, and shares copy-paste prompts. The article is partially paywalled, so the closing prompt library is not visible.
main points
unique insights
practical applications
key topics
key insights
learning outcomes
• main points
1
Systematic head-to-head comparison: 7 tools across 9 realistic search scenarios, with each scenario mapped to a named winner
2
Covers diverse, practical use cases (breaking news, person research, product choice, YouTube, academic papers, Reddit mining, regulations, market sizing, travel)
3
Clear trade-off analysis between standard search and Deep Research (speed, source count, output length, need for prompt engineering)
• unique insights
1
Perplexity's YouTube search outperforms Gemini, which is counterintuitive given Google's ecosystem
2
Deep Research offers 50+ sources and 20+ page outputs but can take 15-25 minutes, and longer output does not imply better quality
3
Tool choice should be matched to task type: different models win different query categories rather than one universal winner
• practical applications
Readers get an actionable tool-selection map for AI-assisted research, reusable prompt templates for specific search types, and guidance on when to use standard search versus Deep Research. Value is limited for readers without access to the paywalled prompt library and rests on a single-author, non-standardized test with a small sample.
• key topics
1
Comparative evaluation of AI search tools (ChatGPT, Perplexity, Grok, DeepSeek, Claude, Gemini, Kimi)
2
Standard AI search versus Deep Research modes
3
Prompt patterns for specific search tasks such as source citation, timestamps, and domain-restricted searches
• key insights
1
Empirical, task-by-task tool comparison rather than generic feature lists
2
Actionable prompt templates with constraints (source counts, citation format, date ranges, Reddit-only filters)
3
Explicit guidance on picking between fast search and slow deep research based on the task
• learning outcomes
1
Identify which AI search tools tend to perform best for different query types such as news synthesis, YouTube, Reddit mining, and regulatory lookups
2
Understand the trade-offs between standard AI search and Deep Research modes and when to use each
3
Apply structured, constraint-based prompts (source counts, citation format, date filters, domain restrictions) to improve AI search output
“ Introduction: Why Testing AI Search Tools Matters
Many people use AI for two main tasks: searching for information and writing content. With dozens of AI search tools now available, choosing the right one can be confusing. This guide tests the leading options side by side to reveal which AI tool is best for online search and how to use it effectively.
“ The Testing Method and Tools Compared
The author ran nine identical search queries through seven AI tools, totaling sixty-three queries. The tools were ChatGPT, the original AI assistant; Perplexity, built specifically for search; Grok from xAI, which has access to Twitter; DeepSeek, the Chinese model that recently received an upgrade; Claude, which launched Sonnet 4.5; Gemini, Google's assistant known for image editing; and Kimi, another Chinese tool recognized for relatively low bias. Each query was designed to reflect a realistic use case, from breaking news to academic research and travel planning.
“ Why Perplexity Wins Overall
Across the nine case studies, Perplexity was the clear overall winner. It located the most information about the author, outperformed even Gemini at finding YouTube video summaries, and searched through 51 sources quickly, far faster than ChatGPT. Its answers were well formatted and followed instructions such as restricting results to Reddit. In one case, it was the only tool to provide the exact address for submitting papers, illustrating its practical value.
“ Category-by-Category Winners Across Nine Case Studies
No single tool dominated every category. Performance depended on the type of task, so matching the tool to the query yields the best results. The sections below summarize the winner for each case study.
“ Breaking News, Person Research, and Product Choices
For breaking news synthesis, summarizing events from the last 72 hours with timestamps and cited sources, DeepSeek produced the most concise and complete answer. For building a current public profile of a person, Perplexity delivered remarkable depth with verifiable links. For high-intent product choices, such as finding a budget ping pong racket, ChatGPT offered the most accurate pricing, Kimi surfaced the most reviews, and Grok provided the best overall recommendation.
“ YouTube Search, Academic Papers, and Reddit Mining
Perplexity was the strongest choice for locating authoritative YouTube videos with timestamps and credibility notes. For finding highly cited academic papers on prompt engineering and hallucinations, all tools performed equally well. For extracting user pain points from Reddit, Perplexity and Grok were the most effective, grouping complaints into clear themes with linked threads.
“ Regulations, Market Sizing, and Travel Planning
For regulatory lookups, Perplexity provided the actual application form and where to submit it, making it especially practical. For estimating market size across TAM, SAM, and SOM with explicit assumptions, Claude and Grok performed best, with Claude holding a slight edge. For a constrained three-day travel itinerary, DeepSeek produced a helpful diagram while ChatGPT delivered the most precise plan.
“ Standard Search vs. Deep Research: How to Choose
Standard AI search returns quick answers from a limited set of sources and works well for most everyday questions. Deep Research goes further, typically consulting more than fifty and sometimes over one hundred sources and producing very long outputs. Its advantages include requiring less prompt engineering and greater coverage. Its drawbacks are long processing times, often fifteen to twenty-five minutes, the need to read lengthy results carefully, and the fact that it still does not search the entire web.
“ Recommended Setup for Deep Research
The author's preferred approaches are the Research tab in Perplexity paired with Claude Sonnet 4.5 Thinking, which is a strong choice overall. ChatGPT's deep research mode is also effective and requires no additional subscription. Claude's research feature, using the new Sonnet 4.5 model, offers a good balance of concise output and deep searching.
“ Copy-and-Paste Prompts for AI Search
Because Perplexity is the strongest general-purpose search tool, the article provides a set of tested prompts that users can copy directly into it. These prompts are designed to deliver structured, cited, and verifiable answers across research tasks. The full prompt collection is available to paid subscribers of the original publication.
“ Conclusion
Perplexity is the best all-around AI tool for searching online, with ChatGPT, DeepSeek, Claude, and Grok each excelling in specific situations. Choosing the right tool for each task, and deciding between standard search and deep research, will produce more reliable and useful results.
We use cookies that are essential for our site to work. To improve our site, we would like to use additional cookies to help us understand how visitors use it, measure traffic to our site from social media platforms and to personalise your experience. Some of the cookies that we use are provided by third parties. To accept all cookies click ‘Accept’. To reject all optional cookies click ‘Reject’.
Comment(0)