Gemini Scans Photos and Gmail for Personalised Answers, But Has Limits

Google has taken its AI assistant, Gemini, a step deeper into personalisation. The company has announced a new feature called “personal intelligence,” allowing Gemini to scan a user’s connected Google apps — including Gmail and Google Photos — to deliver more tailored and context-aware responses. While the update is designed to make the chatbot more useful in everyday situations, Google has acknowledged that the system may struggle with sensitive or complex personal matters, particularly relationship-related issues.

With personalised intelligence enabled, Gemini can retrieve specific details from a user’s own data to answer questions more accurately. For instance, instead of manually searching through old emails or scrolling endlessly through photo libraries, users can simply ask Gemini for information such as a vehicle’s license plate number, tyre size, or travel details. The AI then references relevant emails or images to provide the answer.

The feature also integrates data across multiple Google services, including Gmail, Photos, YouTube and Search, bringing them together into a single conversational interface. According to Google, this allows Gemini to assist with tasks such as planning trips by analysing past travel experiences, suggesting activities based on previous interests, or even recommending games and entertainment drawn from earlier family outings.

However, Google has been transparent about the limitations of this approach. The company admits that Gemini may sometimes overinterpret patterns in a user’s data, leading to assumptions that miss important context. One example shared by Google involves the AI inferring that a user loves baseball simply because they appear frequently in photos taken at a baseball stadium. In reality, those visits may have been family obligations rather than a personal interest.

This limitation becomes even more sensitive when applied to personal relationships. Google notes that if a user has gone through a breakup or divorce, Gemini may still surface memories, suggestions or assumptions based on photos and emails from that past relationship. In such cases, the AI may fail to fully understand emotional nuance. To address this, Google encourages users to correct Gemini directly during conversations by clarifying preferences or situations, helping the system refine its responses.

Privacy remains a major concern when AI tools gain access to personal data, and Google has stressed that safeguards are in place. The company says Gemini does not train on users’ personal images, emails or files. Instead, the AI only references this information temporarily to answer specific questions. Training focuses on prompts and responses, with personal details filtered or obscured to protect user identity.

Gemini also clearly indicates which app it referenced when providing a personalised answer, giving users transparency about the source of the information. If anything seems inaccurate or unclear, users can ask follow-up questions, correct the AI, or adjust connected apps at any time.

Currently, Gemini’s personalised intelligence feature is available in beta to Google AI Pro and AI Ultra subscribers in the United States. Users can enable it by navigating to Gemini’s settings, selecting the personal intelligence option, and choosing which Google apps they want to connect. Access can be modified or revoked at any time, and chat history can be deleted, giving users ongoing control over their data.

As Google continues to push Gemini toward deeper personalisation, the update highlights both the promise and the challenges of AI that understands users more closely — offering convenience and efficiency, while still grappling with the complexities of human relationships.