AI Summary FAQ
AI Summary is designed to streamline the search experience by delivering instant, human-like answers directly on the search results page. Here’s how it works:
When a user enters a query, AI Summary immediately reviews the top three search results, which are identified as the most relevant content. Using advanced AI technology, it extracts key information and generates a concise, accurate summary based on these results. This summary is displayed at the top of the search results page, allowing users to get a direct answer to their question without needing to dig deeper.
The technology behind AI Summary is known as Retrieval-Augmented Generation (RAG), a technique widely recognized for enhancing the accuracy and reliability of AI responses. By leveraging RAG with the latest ChatGPT models from OpenAI, AI Summary is able to provide highly relevant, contextually accurate responses anchored in real data from your site.
Below, you’ll find some frequently asked questions about AI Summary that will help you understand the technology behind this feature.
Didn’t find the answer to your question? Don’t hesitate to contact support and we’ll make sure to provide an answer and add the question to the article!
Where is my information stored when I use AI Summary, and how long is it kept?
When you submit information in AI Summary, it is processed by OpenAI, where it’s stored for a maximum of 30 days solely to check for abuse; OpenAI does not use this data to further train its models. Additionally, your information is stored in our database, which powers the analytics in MyCludo, and the data remains there as long as you are a customer or until you request its deletion. Beyond OpenAI and our database, your information is not stored or processed elsewhere.
Can AI Summary provide answers from documents crawled on a customer’s site?
Yes, AI Summary can return answers from all documents crawled on a customer’s site. However, by default, there’s a 30% boosting applied to files, meaning document-based answers may not always appear prominently unless your query specifically brings up documents as the top results. You can adjust the boosting settings if you want AI Summary to prioritize documents more heavily in response to specific questions.
How does AI Summary avoid hallucinations and give accurate answers?
AI Summary avoids hallucinations by only using content from your website, keeping it focused on real, relevant information. This reduces the chance that it will “hallucinate” or make things up. We use the latest ChatGPT models from OpenAI, which are specifically improved to lower the risk of hallucinations and provide reliable answers. OpenAI updates these models regularly to keep accuracy high, so AI Summary consistently delivers answers you can trust based on your actual content. If there’s any information of utmost importance where accuracy is critical, you can use AI Rules (feature expected to launch early Q1) to manually control responses for specific queries.
Can I customize responses in AI Summary to ensure specific answers for certain topics?
Yes, with our upcoming AI Rules feature, available in early Q1, you’ll have the ability to set custom rules in AI Summary for specific queries. This new feature lets you define keywords or phrases that will trigger predefined responses, ensuring that key topics receive consistent answers. For example, if you set a rule for the keyword “taxes,” AI Summary can always display a specific answer, like “You can see your tax report today,” whenever an end-user asks about taxes, AI Rules gives you more control over the accuracy and consistency of responses, tailored to your needs.
How does AI Summary work to deliver relevant answers?
AI Summary works by leveraging Cludo’s core search backend to retrieve the most relevant results for each query. Here’s an overview of how it functions:
- Search Backend: AI Summary uses the same powerful search backend as other Cludo searches, pulling information from the top search results, much like a standard Search Engine Results Page (SERP). This ensures the answers come from the most relevant content available.
- Retrieval-Augmented Generation (RAG): We use a technique called Retrieval-Augmented Generation (RAG), which is a trusted industry standard for AI-driven responses. AI Summary gathers information from the top results and sends it to ChatGPT, OpenAI’s latest flagship model, which summarizes the key details into a concise answer. By focusing on the top results, AI Summary achieves both high relevance and accuracy in responses.
Extensive experiments confirm that using the top results provides consistently reliable answers. However, we continuously monitor and optimize AI Summary, understanding that different customers and use cases may benefit from varied approaches.
Is the AI Summary feature accessible, and does it comply with WCAG 2.1 AA standards?
Yes, the AI Summary feature is designed with accessibility in mind and conforms to WCAG 2.1 AA standards. Accessibility was a key consideration during its design and build, ensuring that it is accessible to a wide range of users, including those with disabilities.
What is AI Summary Analytics, and how does it help me understand my AI Summary usage?
AI Summary Analytics provides insights into how your AI Summary feature is performing. With the new Generative AI Summary dashboard modules, you can easily monitor general analytics, access a dedicated analytics page, and view an updated account overview to track your usage.
The analytics dashboard gives a clear view of key metrics, such as the total number of searches, success rates, and feedback trends. Additionally, the dedicated AI Summary Analytics page allows you to dive deeper into specific summaries generated for various search terms, helping you understand which topics are most frequently summarized and how users are interacting with them.