On July 23, 2025, Instructure published a press release announcing the name “IgniteAI” for tools in Canvas built with various LLMs (large language models), as well as integrations with partner companies. The following week, Instructure’s head product manager posted more about IgniteAI on their blog. Most of the tools listed in the blog are not yet generally available or are partner integrations. I have gathered information about the tools listed as generally available for all in Canvas: Smart Search, Discussion Summaries, Translations, and Ask Your Data.
Smart Search
Smart Search is the first Canvas tools that can search several types of content in a course from one search query. The search index currently includes Announcements, Assignment descriptions, Discussion topic posts, and Pages. We piloted Smart Search in a handful of courses in Spring 2025. At this time, Smart Search is disabled by default in official academic courses, but teachers may enable it now in a course by (1) going to Settings > Feature options and enabling the Smart Search option, then (2) going to Settings > Navigation, enabling Smart Search in the course menu and clicking Save. Smart Search is enabled by default in sandbox courses.
When enabled, all course users can use Smart Search. Its base LLM is Cohere Embed Multilingual running on AWS Bedrock. Canvas’s own database stores the search index of course contents after the model creates it. The model itself does not store or train on the data it receives about course contents.
Here are links for further reading about Smart Search:
Discussion Summaries
Discussion Summaries allows teachers to request summaries of replies to discussion topics (up to 25 summaries per day) in their courses. At this time, it locked and disabled in our Canvas system. Canvas system admins in Teaching & Learning Technologies have tested it in our beta environment, but it is still under development. We are not currently planning to unlock the feature until it is more thoroughly tested.
Discussion Summaries cannot be generated by students. Claude 3 Haiku running on AWS Bedrock is the base LLM for Discussion Summaries. The model does not store or train on discussion data it receives to generate summaries. Canvas’s own database stores summaries.
For further information about Discussion Summaries:
Translations
Translations features are scheduled to become available in production Canvas environments after the September 20, 2025 release October 18, 2025 release. We do not currently have a plan to enable these features due to the need for more testing.
Translations features in Discussions, Announcements, and Conversations/Inbox messages will allow users to quickly translate text between the following languages: Catalan, Chinese (Simplified), Dutch, English, French, German, Portuguese (Brazil), Russian, Spanish, Swedish. Any course user will be able to generate translations in Discussions and Announcements. In Inbox messages, it will be the sender’s prerogative.
Translations in Discussions and Announcements will be generated by Claude 3 Haiku, and translations in Inbox messages will be generated by AWS Translate. After the release was delayed last December, a product manager posted an explanation on their blog indicating that the feature’s original base model would change. At this time, the “Nutrition Facts” page lists Meta M2M-100, the previous base model. Assuming the rest of the information in the “Nutrition Facts” page hold, the models will not store or train on data received to generate translations.
For further reading about Translations:
Ask Your Data (Intelligent Insights)
These features are only available to Canvas system administrators, and they require an additional fee. At Emory, we have not signed up for this product as we built our own Canvas Data warehouse long before this product was available. For further reading, check out my past posts about data.
Partners
We have not tested the partner integrations that are available now. Because Instructure is not building them into Canvas, we would need to receive requests to review them, and Emory would also need an enterprise license with the partner company.
The featured image is an art print entitled
“Parading through the Streets in Single File”
by Félix Edouard Vallotton in 1893.