[c0n5pir4cy]

Honestly it wasn't even having the first capable chat LLM. OpenAI released the first LLM that has interesting characteristics, based on an architecture they created to be good at this task (Generative Pre-trained Tranformers) and then iterated on improving those characteristics - It took them a while before they had a model capable of following instructions well. There was lots of interest in OpenAI when they released GPT-1 and much more when they released GPT-2.

All the while this was going on Google slept - I think the reasons for this are varied and many, stretching into company culture, structure, processes etc. This was after researchers at Google created the Transformer architecture in the first place.

Google seemed to panic suddenly when ChatGPT was released. I think this is when the people who could actually make change realised that ChatGPT could be a threat to Google in a number of ways - including for search market share.