OpenAI rolls out GPT-5.5 Instant with fewer hallucinations and cleaner responses

OpenAI has introduced a new update to its chatbot platform, aiming to address some of the most common frustrations users have reported when interacting with ChatGPT. The latest model, GPT-5.5 Instant, is now being rolled out globally to all users, including those on the free tier.

The update marks a notable shift in how the chatbot delivers responses. Previous versions had been criticised for overly verbose replies, excessive formatting, and a tendency to include unnecessary elements such as emojis or repeated follow-up prompts. In some cases, the system also produced inaccurate or fabricated information, commonly referred to as hallucinations.

According to OpenAI, GPT-5.5 Instant significantly reduces such issues. The company reports a 52.5 per cent drop in hallucinated claims when handling high-stakes queries across medicine, law, and finance. At the same time, improvements have been made in image reasoning and in performance on science and mathematics tasks. The model is also designed to generate responses that are clearer, shorter, and less cluttered.

Improved responses and more natural interaction

One of the key changes in GPT-5.5 Instant is its tone and structure. OpenAI says the model offers “a more natural conversational tone, and better use of the context you’ve already shared when personalisation can help”. This reflects a broader effort to make interactions feel less forced and more aligned with how people naturally communicate.

In practice, this means users are likely to receive more direct answers without unnecessary elaboration. The updated model avoids overexplaining simple queries and reduces the use of long bullet-point lists, which had previously been a frequent feature of responses. It also limits overly enthusiastic language, aiming for a more balanced, neutral style.

Another noticeable change is how the chatbot handles follow-up prompts. Earlier versions often ended responses with multiple suggestions or questions, which some users found distracting. GPT-5.5 Instant adopts a more restrained approach, focusing on delivering the requested information without adding superfluous content.

The update also reflects a shift in OpenAI’s design philosophy. Rather than introducing highly visible or dramatic features, the company appears to be focusing on refining the everyday user experience. The result is a chatbot that feels more consistent and less inclined to overperform in routine interactions.

New memory tools offer greater transparency

Alongside improvements in response quality, GPT-5.5 Instant introduces a new feature called memory sources. This tool provides users with greater visibility into how their data is used to personalise responses, addressing concerns about transparency and control.

When a response is influenced by stored information, users can now select a ‘Sources’ option to view the context behind it. This may include saved memories, previous conversations, uploaded files, or data from connected services such as email accounts. The feature is designed to make it easier to understand why certain information appears in a response.

Users are also given more control over this data. Through a menu attached to each source, they can delete or correct inaccurate information. For example, if the chatbot holds outdated details about a user’s location, it can be updated directly through the interface. This level of control is intended to reduce the risk of persistent errors that could affect future interactions.

Memory sources are being rolled out across ChatGPT’s consumer plans on the web, with mobile support expected to follow. OpenAI notes that the availability of certain personalisation features may vary by region. Enhanced personalisation, including the use of past chats and connected accounts, is currently limited to Plus and Pro subscribers, though broader access is planned in the coming weeks.

Gradual rollout and early user impressions

The release of GPT-5.5 Instant follows the earlier GPT-5.3 Instant model, which will remain accessible to paid users through a configuration menu for the next three months. Notably, OpenAI has skipped a GPT-5.4 Instant version and moved directly to this latest iteration.

Early impressions suggest that the update delivers on its promise of a more streamlined experience. Users have reported that responses feel noticeably more concise, with fewer unnecessary additions. The chatbot appears to focus more closely on the question asked, reducing the sense that it is trying to impress or overextend its answers.

This change may not be immediately striking visually, as it does not introduce dramatic new capabilities or viral features. However, it represents a meaningful improvement for those who regularly rely on AI tools. By reducing friction and simplifying interactions, GPT-5.5 Instant aims to make everyday use of ChatGPT more efficient and predictable.

OpenAI plans to extend the new features to a wider range of users, including those on Free, Go, Business, and Enterprise plans. Mobile support is also expected to be introduced soon, further broadening access to the updated experience.

While the update may appear incremental, its focus on reliability and usability could have a lasting impact. For frequent users, the combination of fewer errors, clearer responses, and improved transparency may prove to be one of the most practical enhancements the platform has introduced to date.