# First AI Assistant Comparison for Journalist Research ## Objective To build a media list for a pitch on a new athlete x brand partnership announcement. Then, to determine which AI-generated research document most effectively identified relevant journalists. ## Results | Model (Research) | Found | Matches | Accuracy | Speed (m) | Efficiency (matches/m) | | ------------------ | -------- | --------- | -------- | --------- | ---------------------- | | Opus 4.1 | **11**\* | **7**\*\* | 60% | 6 | **1.17** | | o3 | 9 | 6 | 67% | 14 | 0.43 | | Perplexity | 7 | 4 | 57% | **5** | 0.80 | | ChatGPT Agent | 6 | 5 | **83%** | 15\*\*\* | 0.33 | | ChatGPT 5 Thinking | 7 | 3 | 43% | 10 | 0.30 | \* 10 if excluding 1 duplicate across lists; I won't count unclear prompting against the model \*\* tied w/ o3 at 6 if excluding 1 duplicate across lists \*\*\* I stopped it early ## Analysis For journalist list building, **Claude Opus 4.1 is the most practical solution**, delivering the optimal balance of results accuracy and speed. Opus 4.1 and o3 both identified 6-7 correct journalists (tied for most), but Opus accomplished this in just 6 minutes compared to o3's 14 minutes—making it 2.3x more time-efficient. Additionally, it is worth noting that OpenAI has since deprecated it's o3 model. The key metric (for scale) is matches per minute of research time and cost (TBD). My 50% email open rate (initial pitch to combined list) validates that AI-assisted research can effectively identify relevant media contacts when manually verified. See [[Second AI Assistant Comparison for Journalist Research]] --- ## Prompt ``` I need two lists of journalists (active media contacts; real people): - who’ve covered Olympic gold medalist and public health advocate Gabby Thomas’ medical/academic track. - who've covered athlete–medical device endorsements (w/ links to the pieces for context). (in the past 12 mo.) U.S. only. I want first name, last name, title, domain, and the link referenced (per the scrape). I don't want you to spend ANY extra effort trying to get more details or contact information. Just get the relevance right the first time and I can do the rest. ``` ## Final List of Journalists (Manually Verified) A total of 19 journalists were identified as the final outreach targets, including: 1. Megan Holohan 2. Kate Murphy 3. Stephanie Apstein 4. Rose Minutaglio 5. Christen Johnson 6. Andy Frye 7. Stacy Jackson 8. Taylor Dutch 9. Samantha Dorisca 10. Sam Brief 11. Matthew Geagan 12. Greg Hyatt 13. Mary Kekatos 14. Korin Miller 15. Marissa Parra 16. Matthew Ryan 17. Savannah Sellers 18. Beth Snyder Bulik 19. Joe Lemire ## Model Responses Reviewed 1. [Claude Opus 4.1 research](https://docs.google.com/document/d/1Mo-HyzKK0V6KEDuVY_9sL5SqQpN2eFagMlXDb7zSy7c/edit?usp=sharing) 2. [OpenAI o3 research](https://docs.google.com/document/d/1DKUFtIXzb5JaYrOTdjHhD44_A8Jk4jTlK1-ZyY_19oI/edit?usp=sharing) (model deprecated) 3. [ChatGPT Agent research](https://docs.google.com/document/d/15OMRJWanc_7jUH_Mr5zIH3sLib_L0qxQFy8gwTowdw4/edit?usp=sharing) 4. [Perplexity research](https://docs.google.com/document/d/13Hzk7fvFzPRNsjxiM-NOklJCzAua0zldpINqXmyAoIA/edit?usp=sharing) 5. [ChatGPT5 Thinking research](https://docs.google.com/document/d/1Gaid5ddceAox-H-LcMfKDlWQn0EHXHG1ffm4MuoF_dQ/edit?usp=sharing)