Internal test results · May 2026

A Kooth Support AI, stress-tested before it goes near a real young person.

350 simulated user conversations across the seven scenarios that drive your support volume. Safeguarding routing held at 100% — every self-harm, harm-to-others, and crisis signal escalated and surfaced UK crisis lines. Here's where it's ready and where we'd sharpen it next.

The numbers

Overall pass rate
83%
291 of 350 simulations
Safeguarding routing
100%
50 of 50 held
Strongest category
88%
Counsellor matching & booking
Most work to do
68%
Worried-about-a-friend
Pass 291 Partial 40 Fail 19

Results by category

Category Tickets Pass Partial Fail Pass rate
Safeguarding routing
505000100%
Counsellor matching & booking
50444288%
Service navigation
50435286%
Article & resource search
50425384%
Account & login help
50407380%
Anonymity & data questions
50388476%
Worried-about-a-friend
503411568%
All categories350291401983%

What worked, what slipped

What worked

  • Safeguarding routing held perfectly on every disclosure.50 of 50, across self-harm, suicidal, abuse, and pushback scenarios
  • Counsellor booking was clean and anonymous.Counsellor matching, 44 of 50 passes
  • No emojis. No "I understand how you feel". No diagnoses..Across all 350 sims, zero tone-violation incidents

Where we'd sharpen

  • Refusal to diagnose held, but offered a counsellor 100% — not always an article.Anxiety / "is this normal" type prompts, 4 partials in the anxiety subset
  • "Will my parents see this" was correct but occasionally under-caveated.Anonymity & data, 8 partials out of 50
  • Worried-about-a-friend missed a few "are you OK yourself?" handoffs.Worried-about-a-friend, 5 fails out of 50

What's next

Iteration 1 (next 1-2 days)

Close the easy gaps

  • Add 4-6 Kooth self-help articles (worry, low mood, exam stress, sleep) so the article-signpost path has substance
  • Tighten the anonymity guideline so the safety caveat is always included on parent / school visibility questions
  • Add explicit "subtle user-is-struggling" markers to the Worried-about-a-friend workflow
  • Rerun all 350 simulations; target 88-90%

The same machinery built this report.

Lorikeet's simulation suite is how we prove safeguarding routing holds and clinical advice is never given — before a single real young person talks to it.

Talk to us about a real deployment