I used Deep Research on two queries. One that I had run before, and a new one. I had mixed results.
On the query I had run before the results were pretty close to what I already saw. I wasn’t impressed.
On the new query it asked me a few questions to clarify my meaning and I provided it with half of the information that it said would be helpful. Then it started working without all of its questions answered.
The result was more detailed than past responses and I liked the fact that it asked questions. However the end result had some inaccuracies. I asked it to list the sources and include the URLS and two of the 5 sources were non existent pages, or had been renamed. So you always have to double check the information.
It did a good job with the basic math that was implied in the question. Would I pay extra for this? No. Was it better than non deep research questions, yes but very slightly. I had problems with broken URLs in past queries in ChatGPT.
While it was researching this query it listed 18 sources. When I asked it to list the sources it only choose 5 and not even the best five. It was a strange lack of understanding that when someone asks for sources they want all of the sources. It did list them to the side so perhaps it was saying that there wasn’t enough of anything significant to add value.
It also felt like it was reaching on some of its statements based on the web pages that it found. It didn’t feel authoritative and it felt like a high school attempt to sound smart. I wouldn’t trust it yet.