Search that finds what your members are actually looking for.
Most course platforms have a search box. Type a keyword, get a list of lessons whose titles contain that word. Useful if you remember the title; useless if you remember the idea.
lernaura's search is built differently. Every lesson, video transcript, PDF, slide deck and community post is indexed as one semantic surface. Members ask in natural language and get the exact moment that answers the question — across every course, every community thread, every uploaded file in your catalogue.
What it actually does
A member of your community types: "how did the instructor explain the difference between churn and contraction?"
What they get back:
- The 90-second clip in your Customer retention course where you explained it, queued to the right timestamp.
- The community thread from three months ago where another member asked the same question and you replied.
- The PDF cheat-sheet you attached to module 6 that defined both terms.
One query. Three artefacts. Each linked, contextualised, and ready to consume in under a minute.
This is what creators describe as the moment the platform "clicks" for their members — when search stops being a way to fail to find things and starts being a way to extract value from content that was previously locked inside your library.
What gets indexed
Everything in your content catalog. Specifically:
Lesson text — written content in lessons, articles, newsletter posts.
Video transcripts — every video uploaded is transcribed via Gladia (France, no training) and indexed by content, with timestamps preserved. More on the AI content pipeline.
PDFs and slide decks — text extracted, embedded, searchable.
Community posts and replies — public and private (within member access permissions).
Live session recordings — once recorded, treated like any other video.
Quiz and assignment content — questions, instructions, and your own model answers if you've provided them.
The indexing happens automatically. Upload a file, it's searchable within minutes.
Where the AI lives
We use Mistral, a France-hosted model provider, with a contractual no-training clause. That means:
Inference runs in the EU. No data crosses the Atlantic.
Your content does not train any third-party model. Not Mistral's, not Anthropic's, not OpenAI's, not anyone's. Contractually backed.
Embeddings are EU-resident. The numerical representations of your content live on EU infrastructure.
Retention is transparent. Embeddings are retained while your account is active, deleted on cancellation alongside primary data.
This matters in two places. For your European audience, it makes the data location story clean. For your B2B buyers, it satisfies the AI question that procurement reviews now ask by default. More on the European angle.
Search is one view into a single AI surface that runs the whole platform — EU-resident, documented call by call. The architecture behind that is the subject of our AI-first, compliance-first page.
Privacy and access controls
Search respects access permissions. A free-tier member doesn't get returns from paid-tier content. A member of one private cohort doesn't see another cohort's discussions. The system is built on the same access controls that gate the rest of your platform.
You can also mark specific files or threads as excluded from indexing — useful for sensitive cohort discussions or test material that shouldn't surface across the library.
What the AI doesn't do
Three things worth being explicit about:
It doesn't grade. Search surfaces content; it does not assess student answers. Grading that affects course completion or certification stays human-driven.
It doesn't decide access. The AI does not gate, restrict, or unlock content based on engagement scores or behavioural inference.
It doesn't impersonate. No "ask the instructor's AI clone" feature. Members get the actual content you've published, surfaced in a useful order — not generated text in your voice.
These aren't limitations we resent. They're choices we make to stay clean under GDPR Article 22 (automated decision-making about EU residents). More on the Article 22 stance.
What it changes for you
Lower support load. The most common questions in any community get answered by the search itself, with the exact source material attached. Repetitive Q&A drops.
Higher course completion. Members who can find what they need finish what they started. Completion rates in long-form courses improve materially when semantic search is the default discovery path.
More valuable archive. A community with three years of posts is mostly worthless on Discord or Slack because no one can find anything. The same community on lernaura compounds in value — every old thread stays surfaceable forever.
Pricing
Included in the 5% + €0.50, all-in. No separate AI tier, no query-volume cap, no per-member pricing — we only earn when you sell. See pricing.