I would like to report a significant degradation in the performance and accuracy of the AI-generated sequence diagrams.
Until recently, your AI functionality worked flawlessly. I’ve been using very clear, consistent prompts, most of which were generated and refined with ChatGPT, and they used to generate precise, correct diagrams immediately on the first try.
However, something seems to have changed in the past few days. I’ve noticed:
-
The AI interface looks different and responds differently to prompts.
-
Instructions that previously worked are now ignored (e.g., “do not use swimlanes” or “label each step as Step 1, Step 2…”).
-
The AI often duplicates actors, misorders steps, or skips key actions.
-
Diagrams are sometimes correct only after 15–20 retries, which never used to be the case.
It seems that either the AI engine or its prompt parser has been modified recently, because this behavior was never a problem before.
This unreliability is starting to impact my workflow significantly, and I'm now actively considering purchasing a separate diagram AI tool to handle sequence process maps more consistently.
Could you please confirm if any updates were made recently to Lucidchart’s AI engine?
And more importantly: how can we restore previous accuracy and behavior?
