Gemini gave fake 2025 sci-fi books and broken Amazon links
Gemini was asked to find the top 5 science fiction books from 2025, add three sentences about each one, and include an Amazon purchase link. The results included book titles that did not appear to be real, and the purchase links did not work.
Several Gemini models were tried, including Pro and Extended Thinking. Extended Thinking did a little better by returning some older books, with 4 out of 5 appearing to exist, but the direct links still failed.
The same request in ChatGPT produced a useful list, explained what the ranking was based on, and included working purchase links. The main issue is that Gemini Pro performed poorly for a task that depends on factual recommendations and reliable links.
Key points
- Gemini returned questionable book titles for a 2025 science fiction recommendation request.
- The Amazon purchase links did not work.
- Multiple Gemini models were tested, including Pro and Extended Thinking.
- Extended Thinking found more real books, but still failed on direct links.
- ChatGPT handled the same request better, with ranking context and working purchase links.
Quick term guide
- Gemini models
- Google's family of AI models.
- extended thinking
- A mode where an AI works through a detailed internal reasoning chain before giving its answer, using extra tokens in the process.
- ChatGPT Pro
- A paid version of ChatGPT from OpenAI.
- Gemini Pro
- A paid version of Google’s Gemini AI service with extra access or features.
- Solo makers
- People who build and launch their own products or services entirely on their own.
- workflows
- The specific order of steps taken to finish a piece of work.
- automation
- A way to make repeated work happen without doing every step by hand.
- verification
- A separate check that confirms whether the agent finished the task correctly.