AI Analyst gets a major upgrade

We’ve introduced significant improvements to AI Analyst, making it more powerful, flexible and efficient than ever before.

Powered by GPT-5.6 Sol

AI Analyst is now powered by GPT-5.6 Sol, OpenAI’s latest reasoning model designed for more accurate analysis, stronger contextual understanding and improved handling of complex, multi-step intelligence tasks.

Compared to the previous model, GPT-5.6 Sol delivers:

• More robust reasoning across complex intelligence scenarios
• Better understanding of long documents and large collections of reports
• Improved synthesis of information from multiple sources
• Higher-quality analytical writing with more consistent conclusions
• Better handling of uploaded documents and multimodal inputs

In addition to producing better analyses, the new model is also more efficient, consuming approximately 27% fewer tokens on average for the same analytical task. This means lower operating costs while delivering higher-quality results.

Start with your own prompt

We’ve redesigned the AI Analyst workflow to give you more control from the very beginning.

Previously, clicking the AI Analyst button immediately started generating a first draft. While convenient, many users would then refine the analysis by providing additional instructions, requiring a second generation.

Now, when you launch AI Analyst, you’ll first be asked whether you’d like to enter a prompt before the analysis begins.

This allows you to:

• Generate analyses tailored to your exact requirements on the first attempt
• Produce more customised intelligence products
• Save time by avoiding unnecessary regeneration
• Reduce token consumption while obtaining better results

Prefer the previous workflow? Simply press Enter without typing anything, and AI Analyst will immediately begin generating the analysis just as before.

Upload supporting documents

AI Analyst now supports document and image uploads directly within the prompt.

Simply drag and drop your files into the prompt window or click the paperclip icon to attach them.

Supported formats include:

• PDF
• JPEG
• PNG
• CSV

This makes it easier than ever to analyse intelligence reports, images, datasets, spreadsheets and other supporting material alongside Hozint reports.

Turn your analysis into structured tables

AI Analyst can now generate tables directly within your analyses, making complex information easier to organise, compare and interpret at a glance.

Simply ask AI Analyst to present relevant information in a table as part of your prompt. Tables can be particularly useful for:

• Comparing events or incidents across locations, dates, actors, impact levels or other variables.
• Country and regional comparisons, such as comparing security conditions, threat levels or recent developments across multiple countries.
• Chronologies and timelines, organising a sequence of events by date, location and key developments.
• Risk assessments, presenting identified threats alongside likelihood, impact, affected assets and potential mitigation measures.
• Actor analysis, comparing organisations or individuals according to their role, capabilities, objectives and activities.
• Trend analysis, summarising changes in incident volumes, categories or geographic patterns over a selected period.
• Executive briefings, condensing key findings, implications and recommended actions into a format that decision-makers can quickly review.

For example, you can simply add “Summarise the key incidents in a table comparing date, location, incident type and impact” to your prompt, and AI Analyst will incorporate the table directly into the analysis.

The result is more structured, readable and decision-ready intelligence, without having to manually reorganise AI Analyst’s output.

Filter and visualise your analyses on the map

You can now filter analyses using the same filters you already use for alerts and reports, making it easier to find relevant intelligence alongside the events you are monitoring.

For example, you can filter analyses by location, category and timeframe.

To filter analyses, simply select the relevant filters and then go to Filters → More Filters → Report Type → Analysis.

Analyses with geolocation metadata will also appear directly on the map, allowing you to visualise your analytical products alongside other Hozint reports and better understand their geographic context.

To make an analysis searchable and filterable, you first need to add its metadata. Open and edit the analysis, then complete the Metadata section at the bottom of the analysis. Only analyses with the relevant metadata can be retrieved using filters or displayed geographically on the map.
 

Last but not least, the left sidebar used to display only the 25 most recent analyses. It now shows your complete analysis history, making it much easier to revisit previous work, continue ongoing investigations and access older analyses whenever needed.

This turns your growing collection of AI Analyst outputs into a searchable and geospatially integrated intelligence repository, making previous analyses easier to retrieve, compare and reuse.

Tip: If you don’t see your saved analyses on the left sidebar anymore, go to Filters → More Filters → Report Type → select Analysis and then Filters → Timeframe  → remove “Last 30 minutes”

Article clustering: Less noise. Better context. Faster situational awareness.

We have released a major update to the Hozint platform, introducing one of our most requested features: article clustering.

The update also includes source credibility filters and improved AI-generated summaries. 
 

Turning hundreds of articles into one consolidated report

Previously, Hozint AI generated a separate report for each source covering an event. During major incidents, dozens or even hundreds of media outlets could report on the same development, creating unnecessary duplication, information overload and alert fatigue.

You can now activate Article Clustering, a noise-cancelling feature that aggregates articles covering the same event into a single report. Go to Settings > Switch Article Clustering on.

Each clustered report includes:

• The report’s details: Title, Summary, Metadata, Map;
• Real-time source updates as new articles about the event are identified
• A dedicated Media tab where you can view related images and videos

This makes it easier to follow developing events without navigating through hundreds of individual reports.