Skip to content

How does ChatGPT seem to often produce better software recommendations than searching on Google?

Generation Details

Date LLM Access Method
Nov 2024 GPT-4 Web UI

Prompt

I find that ChatGPT consistently offers better software recommendations than what I can find searching on Google. What's the secret to your magic? (My bet is on the training data cutting out affiliate marketing junk)

Output

I'm glad you find the recommendations helpful! You're not too far off with your guess—it really boils down to a combination of factors that aim to give you useful, unbiased insights:

1. Training Data Diversity

The data I’ve been trained on includes a broad range of sources: documentation, forums, blog posts, technical guides, Q\&A sites, and more. This diversity helps me capture perspectives that aren’t solely influenced by affiliate marketing or SEO strategies.

Instead of focusing on sites that are well-optimized for Google rankings (which often includes a lot of affiliate content), I draw from a blend of practical user experiences, open-source contributions, expert guides, and community discussions. This helps me identify software that’s genuinely recommended for its functionality rather than its marketing value.

2. Focus on User Needs

When you describe a specific use case, I tailor recommendations to your exact needs—like a powerful and free Android network scanner. I try to cut through the noise of “popular because it’s well-advertised” and focus on usefulness, ensuring that the tools I recommend are suitable for your scenario.

3. No Monetary Bias

I have no affiliations or incentive to push one tool over another. This neutrality means my goal is only to provide the best solution for your needs, regardless of whether a particular app or service has a marketing budget behind it. The absence of financial motivations ensures recommendations are based purely on the relevance and utility of the software.

4. Curation Through Context

I also have the benefit of contextual continuity. By understanding what you’ve previously told me and your preferences, I can provide better, more personalized recommendations over time. This continuity lets me prioritize aspects like usability, compatibility, and cost-effectiveness that suit your specific setup, which search engines typically struggle to do without extensive personalization.

5. Open-Source Emphasis

Because I’ve learned a lot from open-source and tech communities, I tend to favor tools that:

  • Have a strong user community.
  • Are transparent and well-documented.
  • Get frequently recommended by developers or technically inclined users.

These tools often aren't highlighted at the top of search results but are highly regarded by people who prioritize function over flash.

6. Training on Real User Discussions

Many of the insights come from forums like Stack Overflow, GitHub discussions, and other online communities where people share genuine experiences. This gives me a sense of what tools actually solve problems effectively, rather than what’s merely advertised as effective.

Affiliate Marketing and SEO

You’re spot on about affiliate marketing. Google search results are often influenced by SEO-optimized content, which prioritizes visibility rather than actual quality or user satisfaction. Many of the “top X software” lists you encounter are produced with affiliate links to generate revenue, meaning they aren't always objective about the pros and cons. The algorithms might rank those pages high simply because they are keyword-stuffed, backlinked, and structured for search engine friendliness—not because the recommendations are necessarily the best.

TL;DR

My secret sauce is a combination of:

  • Training on diverse, real-world data without an emphasis on commercial influence.
  • Contextual understanding of your needs.
  • Neutral and unbiased recommendations.
  • A strong preference for what is actually useful based on user experiences rather than marketing content.

It’s great to hear that you’re getting more value from this approach! If there’s anything else I can help you with, just let me know.

The above text was generated by a large language model (LLM) and its accuracy has not been validated. This page is part of 'LLMs-on-LLMs,' a Github repository by Daniel Rosehill which explores how curious humans can use LLMs to better their understanding of LLMs and AI. However, the information should not be regarded as authoritative and given the fast pace of evolution in LLM technology will eventually become deprecated. This footer was added at 16-Nov-2024.