Microsoft was reportedly planning a deeper use of OpenAI’s language models in Bing. The idea could change what users see after entering a search: alongside links, Bing might provide a direct answer or extra context. The reported plans raised questions about how the feature would work and how it would handle mistakes.
From links to generated answers
According to The Information, Microsoft was considering using the technology behind ChatGPT to respond to some search queries. The report cited two people with direct knowledge of the plans. They said the feature could provide an answer or explain a topic, rather than leave users with only a list of pages to open.
The report said Microsoft hoped to introduce the feature before the end of March. The timing, however, was a reported plan, not a confirmed launch. The sources did not identify which model Bing would use or explain exactly how it would appear in search results.
That distinction matters because adding generated text changes the role of a search engine. A list of links points people toward sources; a generated response tries to synthesize information into an answer. Microsoft’s reported approach could combine both, giving people a direct response while retaining search suggestions or links for further exploration.
A longer partnership between Microsoft and OpenAI
The companies had worked together since 2019. The Information had reported in October that Microsoft was planning a second investment in OpenAI and paid almost entirely for the computational costs of training the company’s advanced AI models.
The partnership had already led to AI features across Microsoft products. These included GitHub’s CoPilot and an integration of DALL-E 2 into Bing’s Image Creator. Microsoft also sold GPT-3 and other models to business customers.
According to The Information, the original 2019 investment agreement already covered integrating aspects of GPT into Bing. An earlier version of GPT supported automatic search suggestions. The proposed update would represent a larger step, with sources describing more significant features.
Accuracy is a key open question
The model choice was still unclear. The report named GPT-3.5 and ChatGPT as possible candidates, while noting that the sources had not specified Microsoft’s plans. The models could generate complete responses to search queries or suggest related searches.
Either option would put the quality of the output in focus. The source article said it was not known how Microsoft planned to address hallucinations or other errors. If Bing presents generated text as an answer, users need a way to judge whether it is reliable and find supporting information.
The article did not describe a safeguard or verification method. That leaves a practical question at the center of the reported integration: how would Bing distinguish a useful answer from one that sounds convincing but is wrong? The feature’s design would shape how much users could rely on it.
Potential pressure on Google
The reported plan could also affect competition in search. The source article said ChatGPT had become popular quickly after its November 2022 launch, reaching one million users within five days. It also described Google as rumored to view the chatbot as a competitive threat.
Google had its own language technologies. The article identified LaMDA as an alternative to ChatGPT, while noting that Google had held back from publishing or integrating it into core search. It also said Google had used BERT to better understand search queries since 2019 and was developing MUM, a model designed to handle combined image and text searches.
Whether Google would respond to a more visible AI feature in Bing was uncertain. The source’s account leaves the outcome dependent on details that were not yet public: which model Microsoft chose, what the answers looked like, and how errors were handled. Until those details emerged, the shift remained a reported possibility with potential consequences for how search engines present information.